Karl R. Gegenfurtner Justus-Liebig-Universität, Germany Download paper Play (54min) Download: FLV | MOV | OGG | MP3

The perception of color is a central component of primate vision. Color facilitates object perception and recognition, and plays an important role in scene segmentation and visual memory. Color vision starts with the absorption of light in three different types of light sensitive receptors in the eye, which convert electromagnetic energy into electrical signals, which in turn are transformed into action potentials by a complicated network of cells in the retina. The information is sent to the visual cortex via three independent channels with different chromatic preferences. In the cortex, information from these channels is mixed to enable perception of a large variety of different hues. Furthermore, recent evidence suggests that color analysis and coding cannot be separated from the analysis and coding of other visual attributes such as form and motion. While there are some brain areas that are more sensitive to color than others, color vision emerges through the combined activity of neurons in many different areas.

Marcel P. Lucassen, Theo Gevers, Arjan Gijsenij University of Amsterdam, The Netherlands Download paper Play (20min) Download: FLV | MOV | OGG | MP3

What happens to color emotion responses when texture is added to color samples? To quantify this we performed an experiment in which subjects ordered samples (displayed on a computer monitor) along four scales: Warm-Cool, Masculine-Feminine, Hard-Soft and Heavy-Light. Three sample types were used: uniform color, grayscale textures and color textures. Ten subjects arranged 315 samples (105 per sample type) along each of the four scales. After one week, they repeated the full experiment. The effect of adding texture to color samples is that color remains dominant for the Warm-Cool, Heavy-Light and Masculine-Feminine scale (in order of descending dominance), the importance of texture increases in that same order. The Hard-Soft scale is fully dominated by texture. The average intra-observer variability (between the first and second measurement) was 0.73, 0.66 and 0.65 for the uniform color, grayscale texture and color texture samples, respectively. The average inter-observer variability (between an observer and the other observers) was 0.68, 0.77 and 0.65, respectively. Using some 25,000 observer responses, we derived analytical functions for each sample type and emotion scale (except for the Warm-Cool scale on grayscale textures). These functions predict the group-averaged scale responses from the samples’ color and texture parameters. For uniform color samples, the accuracy of our functions is significantly higher (average adjusted R² = 0.88) than that of functions previously reported. For color texture, the average adjusted R²=0.80.

Sabrina Lachheb, Philippe Colantoni, Éric Dinet (Presented by Alain Trémeau). Laboratoire Hubert Curien, France. Download paper Play (21min) Download: FLV | MOV | OGG | MP3

A set of colours aesthetically pleasant are described as in the language of human visual perception. As this encloses a subjective part, a psychophysical experiment carried out to estimate the perception of colour harmony combinations of paintings with the uniform colour of walls which they are hung.

The experiment, that involved 38 observers, was based on colours built upon a specific colour flow. Participants asked to judge the colour harmony of combinations of of 7 selected paintings with backgrounds uniformly in 3 different ranges of colours—achromatic colours, derived from the global average colour of the considered and tones derived from the complementary of the average colour of the considered painting.

Results demonstrate that the best colour harmony is when the average colour of paintings is used to colour background. The experiment presented in this paper shows that the white colour usually used for walls does not optimize the colour harmony.

Dimitris Mylonas, Lindsay MacDonald London College of Communication, UK Play (18min) Download: FLV | MOV | OGG | MP3

An online colour-naming experiment was designed and developed to determine a broad set of colour words in wide cultural use with their corresponding colour ranges in sRGB and Munsell specifications. The quality of the English dataset was analysed in terms of colour centroids, frequency of names, consistency and response time. The importance of basic colour terms was confirmed while it was also revealed that in free colour naming tasks, the majority of the observers preferred to use non-basic colour terms. The validation of the web-based experimental methodology with previous studies conducted in controlled viewing conditions produced satisfactory results, while a comparison of the 27 most frequent chromatic colour words with a previous web-based experiment showed a remarkable agreement.

Javier Vazquez-Corral1, Graham D. Finlayson2, Maria Vanrell1 1Universitat Autònoma de Barcelona, Spain, 2University of East Anglia, UK Download paper Play (16min) Download: FLV | MOV | OGG | MP3

Understanding how colour is used by the human vision system is a widely studied research field. The field, though quite advanced, still faces important unanswered questions. One of them is the explanation of the unique hues and the assignment of color names. This problem addresses the fact of different perceptual status for different colors.

Recently, Philipona and O’Regan have proposed a biological model that allows to extract the reflection properties of any surface independently of the lighting conditions. These invariant properties are the basis to compute a singularity index that predicts the asymmetries presented in unique hues and basic color categories psychophysical data, therefore is giving a further step in their explanation.

In this paper we build on their formulation and propose a new singularity index. This new formulation equally accounts for the location of the 4 peaks of the World colour survey and has two main advantages. First, it is a simple elegant numerical measure (the Philipona measurement is a rather cumbersome formula). Second, we develop a colour-based explanation for the measure.

Peter Zolliker, Zofia Baranczuk Swiss Federal Laboratory for Materials Testing and Research (EMPA), Switzerland Download paper Play (18min) Download: FLV | MOV | OGG | MP3

Pair comparison methods based on Case V of Thurstone’s Law of Comparative Judgment are widely used to derive interval scales for perceptual image quality. A thorough treatment of the involved statistical errors is often neglected, even though this is the base for computing confidence intervals and other statistical tests. In this paper we show, that consequent error estimation through all steps of the data analysis provides a simple and reliable method to compute confidence intervals. Monte Carlo simulations are used to verify the results and to compare the proposed error estimation with other known methods.

Sérgio M.C. Nascimento, Paulo E.R. Felgueiras, João M.M. Linhares University of Minho, Portugal Download paper Play (24min) Download: FLV | MOV | OGG | MP3

The relationship between the spectral composition of light sources and the visual appearance of rendered scenes is a matter of practical relevance and assumes today particular significance with the advent of light sources of almost arbitrary spectral distribution, like modern LED based lighting. This relationship has only been studied for specific illuminants, like daylights, and systematic studies with other light sources are necessary.
The aim of this work was to address this issue by studying, computationally,
some chromatic effects of metamers of daylight illuminants. For each daylight with correlated color temperature (CCT) in the range 25000 K – 4000 K a large set of metamers was generate using the Schmitt’s elements approach. The metamers set was parameterized by the absolute spectral difference to the equienergy illuminant E and by the number of non-zero spectral bands. The chromatic effects of the metamers were quantified by the CIE color rendering index CRI and by the CIELAB color gamut generated when rendering the Munsell set. It was found that although CRI decreases with ∂, that is, as the illuminant spectrum becomes spectrally more structured, the largest values for the color gamut could be obtained only for large values of ∂. Furthermore, the relationship between color gamut and number of non-zero bands showed that the largest gamuts were obtained with a small number of spectral bands. Thus, spectrally structured metamers produced low CRI but larger color gamuts, a result suggesting that appropriate spectral tuning may be explored in practical illumination when obtaining
large chromatic diversity may be important.

C. Alejandro Párraga, Ramon Baldrich, Maria Vanrell Universitat Autònoma de Barcelona, Spain Download paper Play (17min) Download: FLV | MOV | OGG | MP3

The characterization of trichromatic cameras is usually done in terms of a deviceindependent color space, such as the CIE 1931 XYZ space. This is indeed convenient since it allows the testing of results against colorimetric measures. We have characterized our camera to represent human cone activation by mapping the camera sensor’s (RGB) responses to human (LMS) through a polynomial transformation, which can be “customized” according to the types of scenes we want to represent. Here we present a method to test the accuracy of the camera measures and a study on how the choice of training reflectances for the polynomial may alter the results.

Jutta Willamowski, Frederic Roulland, David B. Martin Xerox Research Centre Europe, France Download paper Play (16min) Download: FLV | MOV | OGG | MP3

In this paper we propose a novel and coherent approach to control and adjust colour reproduction where strict end-to-end colour management cannot be achieved. We first recall the studies and findings that identified the need to reconsider colour management in certain workflows. We then present in detail the Print Mediator system, constituting a first attempt to implement this new approach.

Bernhard Hill RWTH Aachen University, Germany Download paper Play (27min) Download: FLV | MOV | OGG | MP3

A new design of a softproofing system for accurate colour matching is presented in the first part of this paper. In a cabinet, an LCD display is mounted in the viewing plane and illuminated by fluorescent tubes. This arrangement allows for the direct visual comparison of original surface colours with their displayed reproductions face to face. Test colours are placed in cutouts of a mask on top of the screen for colour matching
experiments (colour checker mask) or for calibration purposes. The device provides a ©2010 Society for Imaging xii Science and Technology means for very sensitive proofing of the reproduction algorithm and offers the study of the variability between different observers. The reproduction algorithm applied is based on the spectral specification of all the essential components of the system and allows for fast switching between colour matching functions of different observers considered. Experimental results and studies on the reproduction for different observers and observer metamerism problems are presented in the second part of the paper. Colour matching experiments for a white colour under different illuminants have been performed for a number of different persons. The results show very well reproducible shifts of the experimentally matched colours in the chromaticity diagram, if compared with the original colours for the CIE 1931 standard observer. The results differ for the left and right eye matches for most of the observers. Any person exhibits a typical direction of the shifts like a personal characteristic or a “finger print”, yet, the directions of shifts in the chromaticity diagram are different for different observers and any direction might appear. All the shifts of all persons form a cloud around the original colour in the chromaticity diagram.

Matthias Scheller Lichtenauer1, Hanno Hoffstadt2, Andreas Kraushaar3, Berthold Oberhollenzer3, Peter Zolliker1 1EMPA, Switzerland, 2GMG GmbH & Co. KG, 3Fogra, Germany Download paper Play (18min) Download: FLV | MOV | OGG | MP3

In the graphic arts industry, there is a need to convert colorimetric readings taken on one backing (usually white) to values that would have been measured with a different backing (usually black). We describe and compare different models for such a conversion. Starting from published models using linear scaling, we developed a new nonlinear model for a strongly scattering substrate. Another new model was derived from the Clapper-Yule model, which includes effects of internal reflectances. All these models are applicable in both the spectral and the tristimulus domain.

For calibration, we used measurements of the bare substrate on both backings. We intentionally make only use of the measured spectral or XYZ values, and do not require knowledge of the nominal CMYK values. This is particularly useful for arbitrary patches measured with a stand-alone measurement device. The test data sets consisted of a large set of test prints, originating from digital or conventional printing processes, and covering typical ranges of mass per area. Both new models outperformed linear regression models and the spectral versions always yielded better results than their corresponding versions in tristimulus space.

Céline Villa, Raphaël Labayrade Université de Lyon, France Download paper Play (20min) Download: FLV | MOV | OGG | MP3

To identify people’s preference, psychovisual tests are carried out in virtual environment. Images of scenes used for psychovisual tests are natively High Dynamic Range (HDR) images. However, in order to allow a Low Dynamic Range (LDR) display device to project HDR pictures, their dynamic range must be compressed with a tone-mapping operator (TMO). Thus, before tests are carried out, selection of the most suitable TMO, perceptually speaking, is required to present images faithfully, depending on the aim of the application.

In this paper, three different experimental protocols are proposed for assessing the applications of a display device used for psychovisual tests and selecting the most suitable TMO from six candidates. An additional goal is to give a first idea of the applications that can be addressed by this device. Two subjective experiments were conducted and are presented in the paper. The first one is divided in two steps. The results of three different protocols (two protocols in the first test and the third in the second test) for the identification of the preferred TMO for five different scenes are compared and LDR image defects noticed by observers are highlighted in the paper.

Romain Rossier and Roger D. Hersch Ecole Polytechnique Fédérale de Lausanne, Switzerland Download paper Play (21min) Download: FLV | MOV | OGG | MP3

Different inks may have different mechanical and/or optical properties. Existing Yule-Nielsen modified Neugebauer spectral prediction models assume however that the inks forming a color halftone behave similarly, i.e. that a single n-factor can model the lateral propagation of light within the paper as well as non-uniformities of the ink dot thickness profiles. However, if the inks have very different optical or mechanical properties, each ink may be separately modeled with its specific n-factor. In order to predict the reflection spectrum of such color halftones, we extend the ink spreading enhanced Yule-Nielsen modified spectral Neugebauer (EYNSN) model by calculating for each halftone an optimal n-factor as an average of the ink specific n-factors weighted by a parabolic function of the ink surface coverages. We compare the prediction accuracies of the standard EYNSN model where each halftone is predicted by making use of one global n-factor with the predictions accuracies of the extended EYNSN model where each halftone is predicted with its corresponding optimal n-factor derived from the individual ink-specific n-factors. For inks having very different optical and/or mechanical properties, we observe an improvement of the prediction accuracies.

Alexander D. Logvinenko Glasgow Caledonian University, UK Download paper Play (58min) Download: FLV | MOV | OGG | MP3

Colorimetry can predict which lights will look alike. Such lights are called metameric. Two lights are known to be metameric if they have the same tri-stimulus values. Using the tri-stimulus values as the Cartesian coordinates one can represent light colours as points in a 3D space (referred to as the colorimetric space). All the light colours make a tri-dimensional manifold which can be represented as a circular cone in the colorimetric space. Furthermore, colorimetry can also predict which reflecting objects illuminated by the same light will look alike: those which reflect metameric lights. All the object colours can be represented as a closed solid inscribed in the light colour cone provided the illumination is fixed. However, when there are multiple illuminants the reflected light metamerism does not guarantee that the reflecting objects will look identical (referred to as the colour equivalence). In this paper three axioms are presented that allow the derivation of colour equivalence from metamerism. The colour of a reflecting object under various illuminations is shown to be specified by six numbers (referred to as its six-stimulus values). Using the six-stimulus values one can represent the colours of all the reflecting objects illuminated by various illuminants as a cone in the 6D space over the 5D ball.

David Gossow, Peter Decker, and Dietrich Paulus University of Koblenz-Landau, Germany Download paper Play (23min) Download: FLV | MOV | OGG | MP3

Automatic extraction of local features from images plays an important role in many computer vision tasks. During the last years, much focus has been put on making the features invariant to geometric transformations such as a rotation and scaling of the image. Recently, some work has been published concerning the integration of color information into the detection and description step of SIFT. In various evaluations, it has been shown that including color information can increase distinctiveness and invariance to photometric transformations caused by illumination changes. In this paper we build on the results from these approaches and apply them to the SURF descriptor, which is advantageous compared to SIFT in terms of speed, making it a perfect candidate for online applications, for example in the field of robotics. Our results show significant improvements concerning the repeatability and destinctiveness of SURF for 3D objects under varying illumination directions.

In contrast to many other evaluations we also determine the accuracy of the orientation assignment and include this into our comparisons.

Aurora Sáez, Carmen Serrano, Begoña Acha University of Seville, Spain Download paper Play (21min) Download: FLV | MOV | OGG | MP3

Although several colour edge detectors have been proposed, all of them utilized colour Euclidean distances in the colour space to measure colour differences. In this paper a set of colour edge detection algorithms based on colour visual perception is proposed. These algorithms are a combination of vector edge detector, in uniform colour spaces and using perceptual colour difference equations. A comparative study has been performed. To evaluate the detectors performance a set of synthetics images is generated and various measures are used. It can be concluded that detectors based on CIEDE2000 colour difference equation are better regarding the correlation with colour visual perception.

David Rojas Vigo, Joost van de Weijer, Theo Gevers Universitat Autonoma de Barcelona, Spain Download paper Play (21min) Download: FLV | MOV | OGG | MP3

State of the art methods for image matching, content-based retrieval and recognition use local features. Most of these still exploit only the luminance information for detection. The color saliency boosting algorithm has provided an efficient method to exploit the saliency of color edges based on information theory. However, during the design of this algorithm, some issues were not addressed in depth: (1) The method has ignored the underlying distribution of derivatives in natural images. (2) The dependence of information content in color-boosted edges on its spatial derivatives has not been quantitatively established. (3) To evaluate luminance and color contributions to saliency of edges, a parameter gradually balancing both contributions is required.

We introduce a novel algorithm, based on the principles of independent component analysis, which models the first order derivatives of color natural images by a generalized Gaussian distribution. Furthermore, using this probability model we show that for images with a Laplacian distribution, which is a particular case of generalized Gaussian distribution, the magnitudes of color-boosted edges reflect their corresponding information content. In order to evaluate the impact of color edge saliency in real world applications, we introduce an extension of the Laplacian-of-Gaussian detector to color, and the performance for image matching is evaluated. Our experiments show that our approach provides more discriminative regions in comparison with the original detector.

Cindy Torres and Alain Clément Université d’Angers, France Download paper Play (17min) Download: FLV | MOV | OGG | MP3

A new unsupervised vectorial segmentation method developed for color images is presented. Both spatial and color information have been used during the classification process of the pixels. To overcome the problem of memory space associated with multidimensional histograms analysis and avoid a color requantization, this method is CGIV 2010 and MCS’10 Final Program and Proceedings xv based on a multidimensional compact histogram and an original compact spatial neighborhood probability matrix. The multidimensional compact histogram allows a drastic reduction of memory space necessary for coding color histograms without any data loss. Leaning upon the compact histogram, a spatial neighborhood probability matrix has been computed. It contains all non-negative probabilities of spatial connectivity between pixel colors and allows a spatial distance inside a variable size neighborhood to be defined. In an unsupervised histogram analysis classification process, two phases are classically distinguished: (i) a learning process during which histogram modes are identified and (ii) a second step called the decision step in which a full partition of the colorimetric space is carried out according the previously defined classes. In a standard colorimetric approach, in the second step, a colorimetric distance like Euclidean or Mahalanobis is used. We introduced here a spatio-colorimetric distance taking into account the information of pixel neighborhood colors. This distance is defined as a weighed mixture between a colorimetric distance and the spatial distance calculated from the spatial neighborhood probability matrix. The vectorial classification method is based on previously presented principles, achieving a hierarchical analysis of the color histogram using a 3D-connected components labelling.

Aldo Maalouf, Mohamed-Chaker Larabi University of Poitiers, France Download paper Play (20min) Download: FLV | MOV | OGG | MP3

In this work, a two-step technique for constructing a super-resolution (SR) image from a single multi-valued low-resolution (LR) input image is proposed. The problem of SR is treated from the perspective of image geometry-oriented interpolation. The first step consists of computing the image geometry of the LR image by using the grouplet transform. Having well represented the geometry of each color channel in the LR image, we propose a grouplet-based structure tensor whose role is to couple the geometrical information of the different image color components. In a second step, a functional is defined on the multispectral geometry defined by this structure tensor. The minimization of this functional insures the synthesize of the SR image. The proposed super-resolution algorithm outperforms the state-of-the-art methods in terms of visual quality of the interpolated image.

Mihai Ivanovici1, Nöel Richard2, Christine Fernandez-Maloigne2 1Transylvania University, Romania, 2University of Poitiers, France Download paper Play (19min) Download: FLV | MOV | OGG | MP3

Fractal dimension and lacunarity are two fractal measures widely used for image analysis, segmentation and indexation. In this paper, we show how these two fractal features are able to capture several aspects that characterize the degradation of the video signal, based on the fact that the quality perceived is directly proportional to the fractal complexity of an image. Thus, we demonstrate that the fractal dimension and lacunarity can be used to objectively assess the quality of the video signal and how they can be used as metrics for the user-perceived video quality degradation for an MPEG-4 streaming application.

Unfortunately, all the existing approaches are defined only for binary and greyscale images. Based on the probabilistic algorithm for the estimation of the fractal dimension and computation of lacunarity, we propose a colour approach that makes possible the analysis of the complexity in the RGB colour space of any colour image. We discuss our experimental results and then draw the conclusions.

Aram Danielyan, Alessandro Foi, Vladimir Katkovnik, Karen Egiazarian Tampere University of Technology, Finland Download paper Play (15min) Download: FLV | MOV | OGG | MP3

We propose a new algorithm for multispectral image denoising. The algorithm is based on the state-of-the-art Block Matching 3-D filter. For each “reference” 3-D block of multispectral data (sub-array of pixels from spatial and spectral locations) we find similar 3-D blocks using block matching and group them together to form a set of 4-D groups of pixels in spatial (2-D), spectral (1-D) and “temporal matched” (1-D) directions. Each of these groups is transformed using 4-D separable transforms formed by a fixed 2-D transform in spatial coordinates, a fixed 1-D transform in “temporal” coordinate, and 1-D PCA transform in spectral coordinates. Denoising is performed by shrinking these 4-D spectral components, applying an inverse 4-D transform to obtain estimates for all 4-D blocks and aggregating all estimates together. The effectiveness of the proposed approach is demonstrated on the denoising of real images captured with multispectral camera.

Jaime Moreno, Xavier Otazu, Maria Vanrell Universitat Autònoma de Barcelona, Spain Download paper Play (15min) Download: FLV | MOV | OGG | MP3

The aim of this work is to explain how to apply perceptual concepts to define a perceptual pre-quantizer and to improve JPEG2000 compressor. The approach consists in quantizing wavelet transform coefficients using some of the human visual system behavior properties. Noise is fatal to image compression performance, because it can be both annoying for the observer and consumes excessive bandwidth when the imagery is transmitted. Perceptual pre-quantization reduces unperceivable details and thus improve both visual impression and transmission properties. The comparison between JPEG2000 without and with perceptual pre-quantization shows that the latter is not favorable in PSNR, but the recovered image is more compressed at the same or even better visual quality measured with a weighted PSNR. Perceptual criteria were taken from the CIWaM (Chromatic Induction Wavelet Model).

Toko Kohei1, Jinhui Chao1, Reiner Lenz2 1Chuo University, Japan, 2Linköping University, Sweden Download paper Play (23min) Download: FLV | MOV | OGG | MP3

In this paper we discuss the role of curvature in the context of color spaces. Curvature is a differential geometric property of color spaces that has attracted less attention than other properties like the metric or geodesics. In this paper we argue that the curvature of a color space is important since curvature properties are essential in the construction of color coordinate systems. Only color spaces with negative or zero curvature everywhere allow the construction of Munsell-like coordinates with geodesics, shortest paths between two colors, that never intersect. In differential geometry such coordinate systems are known as Riemann coordinates and they are generalizations of the wellknown polar coordinates.

We investigate the properties of two measurement sets of just-noticeable-difference (jnd) ellipses and color coordinate systems constructed from them. We illustrate the role of curvature by investigating Riemann normal coordinates in CIELUV and CIELAB spaces. An algorithsm is also shown to build multi-patch Riemann coordinates for spaces with the positive curvature.

Gerard van Dalen, Faisal Osman, Aat Don Unilever Research and Development, The Netherlands Download paper Play (17min) Download: FLV | MOV | OGG | MP3

Within Food R&D scientists, chefs and engineers aim to develop natural fruit and vegetable based soups with optimal flavour, texture, appearance and health benefits whilst maintaining product safety. The colour and appearance has a major influenceon the perceived quality (e.g. nutrition and freshness). This paper describes a methodfor the determination of the colour of soups containing vegetable particles. Digitalimages were made under controlled lighting using a DigiEye imaging system. It allowsdocumenting the appearance by making colorimetrically accurate images which aresuitable for the measurement of colour uniformity, size and shape. The uniformity of thelight under diffuse and directional illumination was investigated using different targetsand soup samples. For accurate colour analysis the images have to be corrected fornon-uniform illumination and the colour has to be measured at a fixed location outsidethe centre of the lighting cabinet. Directional illumination (angled light with additionalmirrors) was used to introduce gloss resulting in images matching the actual appearanceof soups and vegetable particles very closely. Imaging glossy soups under diffuseillumination resulted in dull images with dark spots. No significant difference was foundbetween the colours of soups analysed under diffuse or directional illumination. Underdirectional illumination a better repeatability was observed. Additionally, for 36 differentsoups, the measured results from the DigiEye system were compared to two differentcolorimeters (0°/45° and diffuse/0° geometry). Linear relations were found betweenthe CIE Lab values measured by the DigiEye system and those measured by two differentcolorimeters. Best correlations were obtained between DigiEye and 0°/45° colorimeter(r2=0.980-0.996). The short term precision of the DigiEye system is somewhat betterthan those of the colorimeter.

Magnus Neuman1, Per Edström1, Mattias Andersson1, Ludovic Coppel1,2, Ole Norberg1 1Mid Sweden University, 2Inneventia AB, Sweden Download paper Play (14min) Download: FLV | MOV | OGG | MP3

The angular variations of color of a set of paper samples are experimentally assessed using goniophotometric measurements. The corresponding simulations are done using a radiative transfer based simulation tool, thus considering only the contribution of bulk scattering to the reflectance. It is seen that measurements and simulations agree and display the same characteristics, with the lightness increasing and the chroma decreasing as the observation polar angle increases. The decrease in chroma is larger the more dye the paper contains. Based on previous results about anisotropic reflectance from turbid media these findings are explained. The relative reflectance in large polar angles of wavelengths with strong absorption is higher than that of wavelengths with low absorption. This leads to a loss of chroma and color information in these angles. The increase in lightness is a result of the anisotropy affecting all wavelengths equally, which is the case for transmitting media and obliquely incident illumination. The only CGIV 2010 and MCS’10 Final Program and Proceedings xxiii case with no color variations of this kind is when a non-absorbing, non-transmitting medium is illuminated diffusely. The measured and simulated color differences are clearly large, and it is an open issue how angle resolved color should be handled in standard color calculations.

Abhijit Sarkar1,2, Laurent Blondé1, Patrick Le Callet2, Florent Autrusseau2, Patrick Morvan1, Jürgen Stauder1 1Technicolor Research, 2IRCCyN-IVC, France Download paper Play (22min) Download: FLV | MOV | OGG | MP3

Various recent studies have shown that observer variability can be a significant issue in modern display colorimetry, since narrow-band primaries are often used to achieve wider color gamuts. As far as industrial applications are concerned, past works on various aspects of observer variability and metamerism have mostly focused on crossmedia color matching, an application context that is different from color matching on two displays, both in terms of human visual performance and the application requirements. In this paper, we report a set of three preliminary color matching experiments using a studio Cathode Ray Tube (CRT) display with broadband primaries, and a modern wide-color gamut Liquid Crystal Display (LCD) with narrow-band primaries, with and without surround. Two principal goals of these pilot tests are to validate the experimental protocol, and to obtain a first set of metameric data of display color matches under different viewing conditions. In this paper, various experimental design considerations leading to the current test setup are discussed, and the results from the pilot tests are presented. We confirm the validity of our test setup, and show that the average color matches predicted by the 1964 CIE 10° standard observer, although acceptable as average matches, can often be significantly and unacceptably different from individual observer color matches. The mean, maximum and the 90th percentile values of the standard observer-predicted color difference of individual observer color matches were 1.4, 3.3 and 2.6 ΔE*00 respectively.

Renzo Shamey1, David Hinks1, Manuel Melgosa2, M. Ronnier Luo3, Guihua Cui3, Rafael Huertas2, Lina Cárdenas1, and Seung Geol Lee1 1North Carolina State University, USA, 2University of Granada, Spain, 3University of Leeds, UK Download paper Play (23min) Download: FLV | MOV | OGG | MP3

We previously reported the performance of four color difference equations around the CIE 1978 blue color center (NCSU-B1) using various statistical measures. In this study we employed the standardized residual sum of squares (STRESS) index to test the performance of twelve color-difference formulae using two experimental NCSU datasets. The first dataset (NCSU-B1) included 66 sample pairs around the CIE 1978 blue color center and the second dataset (NCSU- 2) contained 69 sample pairs around 13 color centers. In the first dataset 26 observers made a total of 5148 assessments of sample pairs with small color differences (ΔE*ab<5) while the second dataset involved 20,700 assessments by 100 observers from four different geographical regions of the world (25 in each region). Each pair in both sets was assessed by each color normal observer in three separate sittings on separate days and the average of assessments was calculated. For the samples in the first dataset a custom AATCC standard gray scale was employed to assess the magnitude of difference between colored samples. A third-degree polynomial equation was used to convert gray scale ratings to visual differences (ΔV). In the second study a novel perceptually linear gray scale was developed and a linear function was used to obtain visual differences. Based on the analysis of STRESS index results the DIN99d equation gave the best results for both datasets, and the CIELAB equation the worst.

Jon Y. Hardeberg Gjøvik University College, Norway Download paper Play (61min) Download: FLV | MOV | OGG | MP3

Quantifying the perceptual difference between original and reproduced (and inevitably modified) color images is currently a key research challenge in the field of color imaging. Such information can be extremely valuable for instance in the development of new equipment and algorithms for color reproduction.

While in many research areas it is common practice to obtain quantitative quality information by the use of perceptual tests, in which the judgments of several human observers are being collected and carefully analyzed statistically, this approach has serious limitations for practical use, in particular because of the time consumption.

Motivated by this, and aided by the ever increasing available knowledge about the mechanisms of the human visual system, the quest for perceptual color image quality metrics that can adequately predict human quality judgments of complex images, has been on for several decades. However, unfortunately, the Holy Grail is yet to be found.

The current paper outlines the state of the art of this field, including benchmarking of existing metrics, presents recent research, and proposes promising areas for further work. Aspects that are covered in particular include new models and metrics for color image quality, and new frameworks for using the metrics to improve color image representation and reproduction algorithms.

Jakkarin Singnoo, Graham D. Finlayson University of East Anglia, UK Download paper Play (23min) Download: FLV | MOV | OGG | MP3

High Dynamic Range (HDR) imaging has become more widespread in consumer imaging in the past few years, due to the emergence of methods for the recovering HDR radiance maps from multiple photographs. In the domain of HDR encoding, the RGBE radiance format (.hdr) is one of the most widely used. However, conventional image editing applications do not always support this encoding and those that do take considerable time to read or write HDR images (compared with more conventional formats) and this hinders workflow productivity. In this paper we propose a simple, fast, and practical framework to extend the conventional 12 and 16-bit/channel integer TIFF gamma-encoded image format for storing such a wide dynamic range. We consider the potential of our framework for the tone-mapping application both by measuring the ΔE S-CIELAB color difference between original and encoded image, and by conducting a psychophysical experiment to evaluate the perceptual image quality of the proposed framework and compare it with an RGBE radiance encoding. The preliminary results show that our encoding frameworks work well for all images of a 65 image dataset, and give equivalent results compared to RGBE radiance formats, while both consuming much less computational cost and removing the need for a separate image coding format. The results suggest that our method, used in the normal tone mapping workflow, is a good candidate for HDR encoding and could easily be integrated with the existing TIFF image library.

Guanqun Cao1, Marius Pedersen1,2, Zofia Baranczuk3 1Gjøvik University College, Norway, 2Óce Print Logic Technologies S.A., France, 3Swiss Federal Laboratory for Materials Testing and Research EMPA, Switzerland Download paper Play (14min) Download: FLV | MOV | OGG | MP3

When an image is reproduced with a device different artificats can occur. These artifacts, if dectectable by observers, will reduce the quality of the image. If these artifacts occur in salient regions (regions of interest) or or if the artifacts introduce salient regions regions they contribute to reduce the quality of the reproduction. In this paper we CGIV 2010 and MCS’10 Final Program and Proceedings xxv propose a novel method for the detection of artifacts based on saliency models. The method is evaluated against a set of gamut mapped images containing the artifacts, which have have been marked by a group of group experts. The results have shown that the proposed metrics are promising to detect the artifacts through the reproduction.

Masayuki Ukishima1,2, Yoshinori Suzuki1, Norimichi Tsumura1, Toshiya Nakaguchi1, Martti Mäkinen2, Shinichi Inoue3, Jussi Parkkinen2 1Chiba University, Japan, 2University of Eastern Finland, Finland, 3Mitsubishi Paper Mills Ltd., Japan Download paper Play (14min) Download: FLV | MOV | OGG | MP3

As the spectral prediction model for color halftone prints using the microscopic measurement, the conventional spectral reflection image (SRIM) is extended by introducing the concept of the conventional spectral Neugebaur Model, and a new production model, the Neugebauer modifies spectral reflection image model (NMSRIM), is proposed. Compared to the SRIM, the NMSRIM abstracts the spatio-spectral transmittance distribution of ink layer using the the limited number of base color functions and the spatial position function for each base color in order to efficiently predict the reflectance of color halftone prints from a small number of measurements. The NMSRIM separately analyzes the mechanical dot gain and the optical dot gain. The NSRIM can predict can predict not only the spectral reflectance but also the microscopic spatial distribution of reflectance. The spatial distribution of reflection of reflectance is related to the appearance of halftone prints. The methods to obtain the parameters of NMSRIM are also proposed. Several parameters are obtained by measurements and the others are obtained by computational estimations. To evaluate the validity of the NMSRIM, the spatio-spectral distrbution of reflectance printed with two inks, cyan and magenta (testing data) is predicted from the measurements of the halftones printed with one ink, the unprinted paper, and the solid prints of ink which are the cyan, magenta and blue (training data), where the blue corresponds to the combination of cyan and magenta inks. The spectral prediction accuracy was significant since the average and maximum values ΔE94 in all samples were 0.66 and 1.30, respectively. We also obtained the interesting results according to the spatial data.

Massimo Fierro, Tae-Hyoung Lee, Yeong-Ho Ha Kyungpook National University, Korea Download paper Play (19min) Download: FLV | MOV | OGG | MP3

HDR image formation and display has been an argument of extreme interest even when digital cameras were not yet consumer products. While recent research in both fields has seen very interesting works, none is really revolutionary, since what goes on behind the scene has been left basically unchanged. In the image formation field in particular, a lot of energy has been spent so to solve the problems that arise when taking multiple exposures: illumination change, camera shake and in-scene movement. In this paper we approach HDR image formation from a different perspective, which tries to solve in one move all the mentioned problems. More specifically, we propose a method that is able to estimate missing exposures for HDR image formation starting from only one under-exposed shot. Estimation is done through artificial neural networks: the development of a mathematical model is a highly desirable, but time consuming task. The results are are very interesting, although not perfect, and suggest that further research might lead to a suitable solution.

Arne Magnus Bakke, Ivar Farup Gjøvik University College, Norway Download paper Play (20min) Download: FLV | MOV | OGG | MP3

Gamut boundary determination is an important step in device characterisation and colour gamut mapping. Many different algorithms for the determination of colour gamuts are proposed in the literature. They vary in accuracy, computational efficiency, and complexity of the resulting triangulated gamut surface. Recently, an algorithm called uniform segment visualization (USV) was developed. The gamut surfaces produced by the USV algorithm is more accurate than the ones produced by the the segment maxima algorithm, while at the same time, they are significantly simpler than the ones produced by the somewhat more accurate modified convex hull. In this paper, we propose a new method. First, an accurate gamut boundary is computed using the modified convex hull. The resulting surface is then simplified using an established mesh decimation technique. This results in surfaces that are significantly more accurate than the ones produced by the USV algorithm at a comparable complexity.

Charles Poynton Simon Fraser University, Canada Download paper Play (20min) Download: FLV | MOV | OGG | MP3

Colour gamut refers to the range of colours that can be reproduced by an imaging system. The definition of gamut is quite clear for displays and for hard-copy printing. Colour image science experts disagree, however, on the definition—or even applicability of the concept—of gamut for cameras. I disagree that there is any meaningful concept of “capture gamut.” In this note, I review trichromacy and metamerism and discuss various gamuts. I conclude that although metamerism is a phenomenon of what I call “31-space,” gamut lives in 3-space. With suitable colour signal processing, a 3-channel camera is capable of acquiring wide gamut images.

Dibakar Raj Pant1,2, Ivar Farup1 1Gjøvik University College, Norway, 2University Jean Monnet, France Download paper Play (19min) Download: FLV | MOV | OGG | MP3

For precision color matching, visual sensitivity to small color difference is an essential factor. Small color differences can be measured by the just noticeable difference (JND) ellipses. The points on the ellipse represent colours that are just noticably different from the colour of the centre point. Mathematically, such an ellipse can be described by a positive definite quadratic differential form, which is also known as the Riemannian metric. In this paper, we propose a method which makes use of the Riemannian metric and Jacobean transformations to transform JND ellipses between different colour spaces. As an example, we compute the JND ellipses of the CIELAB and CIELUV color difference formulae in the xy chromaticity diagram. We also propose a measure for comparing the similarity of a pair of ellipses and use that measure to compare the CIELAB and CIELUV ellipses to two previously established experimental sets of ellipses. The proposed measure takes into account the size, shape and orientation. The technique works by calculating the ratio of the area of the intersection and the area of the union of a pair of ellipses. The method developed can in principle be applied for comparing the performance of any color difference formula and experimentally obtained sets of colour discrimination ellipses.

Rafael Huertas1, Alain Tremeau2, Manuel Melgosa1, Luis Gomez-Robledo1, Guihua Cui3, M. Ronnier Luo3 1Universidad de Granada, Spain, 2Université Jean Monnet, France, 3University of Leeds, UK Download paper Play (23min) Download: FLV | MOV | OGG | MP3

Several colour-difference formulas have been proposed since the last recommendation of CIEDE2000 by the “Commission Internationale de L’Eclairage” (CIE) in 2001. Some of them have been tested using the same dataset used to fit them. Thus, it is of great interest to check the performance of these formulas with new experimental datasets. On the other hand, some previous studies show that many colour-difference formulas perform quite badly in the very small colour difference range of 0 to 1 CIELAB units. This paper pursues these two goals. The colour-difference formulas DIN99d, OSA-GP, OSA-GP Euclidean (OSA-GPE), CAM02-SCD and CAM02-UCS are tested with a new experimental dataset, which has been carried out in the Laboratoire Hubert Curien of Saint Etienne (France) in two different modes, physical metallic samples and virtual samples displayed in a LCD monitor. This new dataset is composed by 390 colour pairs arranged around 16 colour centres with colour differences in the range 0.14 to 2.14 CIELAB units, with an average value of 0.80. In this work only just noticeable differences have been considered from this dataset. The results show a bad performance of all studied colour-difference formulas for just noticeable colour differences, in agreement with previous studies. Further research must be conducted to fit colourdifference formulae to this important range of colour differences.

Henri Kivinen, Mikko Nuutinen, Pirkko Oittinen Aalto University School of Science and Technology, Finland Download paper Play (14min) Download: FLV | MOV | OGG | MP3

Perceptual colour difference in simple colour patches has been extensively studied in the history of colour science. However, these methods are not assumed to be applicable for predicting the perceived colour difference in complex colour patches such as digital images of complex scene. In this work existing metrics that predict the perceived colour difference in digital images of complex scene are studied and compared. Performance evaluation was based on the correlations between values of the metrics and results of subjective tests that were done as a pair comparison, in which fifteen test participants evaluated the subjective colour differences in digital images.

The test image set consisted of eight images each having four versions of distortion generated by applying different ICC profiles. According to results, none of the ©2010 Society for Imaging xxviii Science and Technology metrics were able to predict the perceived colour difference in every test image. The results of iCAM metric had the highest average correlation for all images. However, the scatter of the judgements was very high for two of the images, and if these were excluded from the comparison the Hue-angle was the best performing metric. It was also noteworthy that the performance of the CIELAB colour difference metric was relatively high.

Julie Klein, Johannes Brauers, Til Aach RWTH Aachen University, Germany Download paper Play (21min) Download: FLV | MOV | OGG | MP3

The wavelength-dependency of the refraction indices in optical systems involves chromatic aberrations: one object point is not projected on exactly one image point on the sensor plane, but dispersed depending on its wavelengths in a rainbow-like manner due to the wavelength-dependency. These distortions have already been analyzed for the three broadband color planes red, green and blue of an RGB camera for example. As far as the authors know, no analysis was performed yet for more than three narrow wavelength bands and each color plane was considered separately so far.

In this paper, we describe the measurement of chromatic aberrations for multiple narrowband color channels and extend the models from the literature to characterize these distortions. We then link the parameters obtained for all color channels in order to include the wavelength-dependency into the models of the distortions. This leads to a more general model for the chromatic aberrations, calculating the distortions as a function of the wavelength and the image position. We then compensate the chromatic aberrations using these models and finally estimate their accuracy.

Noriyuki Shimano Kinki University, Japan Download paper Play (20min) Download: FLV | MOV | OGG | MP3

Prior knowledge of the noise present in a color acquisition device is very important for the recovery of spectral reflectance of an object being imaged, since recovery performance is greatly influenced by the noise.

In the previous paper (IEEE Trans. Image Process. 1848 (2006), the author proposed a new model to estimate noise variance of an image acquisition system by assuming the noise variance in each channel is equal and showed this model is very useful to accurately recover a reflectance of an imaged object. This paper describes extended model for the estimation of the covariance matrix the noise present in an image acquisition system without assumption. It is demonstrated that the proposal overfits noise covariance matrix to learning samples and that recovery performance for the test samples is poor with the previous model. However this overfitting means the estimates are correctly performed using the model. The new model is effective in analyzing the present in an image acquisition system.

Oili Kohonen University of Eastern Finland, Finland Download paper Play (21min) Download: FLV | MOV | OGG | MP3

This work examines the spectrum preserving properties of a multi-resolution analysisbased intensity modulation (MRAIM) when used for increasing the spatial resolution of spectral color images. The MRAIM algorithm is originally designed to fuse high-resolution panchromatic images with low-resolution spectral images in order to get high-resolution spectral images for remote sensing applications. Instead of panchromatic images, for which the MRAIM algorithm has originally been designed for, the MRAIM algorithm is implemented to use information from both grayscale and RGB color images. In order to utilize the information of the three channels included in RGB images, two different models are derived and examined. In addition, two kind of scaling factors are used for compensating possible differences between the images acquired at different resolution levels. The resulting high resolution spectral images are compared to the real acquired high resolution spectral images with respect to both maximum and average RMS errors and ΔE*ab color differences under CIE illuminants D65, A, F8 and F11. The used images are acquired by NuanceFX spectral imaging system, which allows the measuring of both spectral and RGB images at different resolution levels at identical geometry.

Óscar Martínez1, Meritxell Vilaseca2, Carles Pizarro2, Èdgar Ferrer2, Monserrat Arjona2, Jaume Pujol2 1Hewlett-Packard España, 2Technical University of Catalonia, Spain Download paper Play (29min) Download: FLV | MOV | OGG | MP3

Multispectral systems allow the spectral characterization of the scene through several acquisition channels with different spectral features. The spectral sampling can be done by using transmittance filters and a white light source or illuminating the scene using light sources with different spectral emission characteristics. Light-emitting diodes based light sources have started to be used in multispectral systems, mainly to develop low-cost devices for the industry. In this study we analyze the spectral power distribution and color variability of white and single color light-emitting diodes relative to the viewing angle, and highlight some aspects that must be taken into account if these light sources want to be used in a multispectral system.

Jukka Antikainen, Markku Hauta-Kasari, Timo Jääskeläinen, Jussi Parkkinen University of Eastern Finland, Finland Download paper Play (15min) Download: FLV | MOV | OGG | MP3

In this study, we implement a fast non-iterative Principal Component Analysis computation for spectral image analysis by utilizing Graphical Processing Unit GPU. PCA inner product computation efficiency between Central Processing Unit CPU and GPU was examined. Performance was tested by using spectral images with different dimensions and different PCA inner product image counts. It will be shown that the GPU implementation provides about seven times faster PCA computation than the optimized CPU ©2010 Society for Imaging xxx Science and Technology version. Difference to the commonly used scientific analysis software Matlab is even higher. When spectral image analysis is needed to make in real-time, CPU does not offer the necessary performance for larger spectral images. Therefore, powerful GPU implementation is needed.

Oh-Seol Kwon, Holly E. Gerhard, Laurence T. Maloney New York University, USA Download paper Play (15min) Download: FLV | MOV | OGG | MP3

Surface reflectance functions (SRFs) and spectral power distributions (SPDs) of illuminants are typically modeled as elements in an N-dimensional linear function subspaces. Each SRF and SPD is represented by an N-vector and the mapping between SRF and SPD functions and an N-dimensional vector assigns N-dimensional “color” codes representing surface and light information. The N basis functions are chosen so that SRFs and SPDs can be accurately reconstructed from their N-dimensional vector codes. Typical rendering applications assume that the resulting mapping is an isomorphism where vector operations of addition, scalar multiplication and component-wise multiplication on the N-vectors can be used to model physical operations such as superposition of lights, light-surface interactions and inter-reflection. When N is small, this implicit isomorphism can fail even though individual SPDs and SRFs can still be accurately reconstructed by the codes. The vector operations do not mirror the physical. However, if the choice of basis functions is restricted to characteristic functions (that take on only the values 0 and 1) then the resulting map between SPDs/SRFs and N-vectors is an isomorphism that preserves the physical operations needed in rendering. The restriction to bases composed of characteristic functions can only reduce the goodness of fit of the linear function subspace to actual surfaces and lights. We will investigate how to select characteristic function bases of any dimension N (number of basis functions) and evaluate how accurately a large set of Munsell color chips can approximated as a function of dimension.