Improving gamut mapping color constancy

The color constancy problem, that is, estimating the color of the scene illuminant from a set of image data recorded under an unknown light, is an important problem in computer vision and digital photography. The gamut mapping approach to color constancy is, to date, one of the most successful solutions to this problem. In this algorithm the set of mappings taking the image colors recorded under an unknown illuminant to the gamut of all colors observed under a standard illuminant is characterized. Then, at a second stage, a single mapping is selected from this feasible set. In the first version of this algorithm Forsyth (1990) mapped sensor values recorded under one illuminant to those recorded under a second, using a three-dimensional (3-D) diagonal matrix. However because the intensity of the scene illuminant cannot be recovered Finlayson (see IEEE Trans. Pattern Anal. Machine Intell. vol.18, no.10, p.1034-38, 1996) modified Forsyth's algorithm to work in a two-dimensional (2-D) chromaticity space and set out to recover only 2-D chromaticity mappings. While the chromaticity mapping overcomes the intensity problem it is not clear that something has not been lost in the process. The first result of this paper is to show that only intensity information is lost. Formally, we prove that the feasible set calculated by Forsyth's original algorithm, projected into 2-D, is the same as the feasible set calculated by the 2-D algorithm. Thus, there is no advantage in using the 3-D algorithm and we can use the simpler, 2-D version of the algorithm to characterize the set of feasible illuminants. Another problem with the chromaticity mapping is that it is perspective in nature and so chromaticities and chromaticity maps are perspectively distorted. Previous work demonstrated that the effects of perspective distortion were serious for the 2-D algorithm. Indeed, in order to select a sensible single mapping from the feasible set this set must first be mapped back up to 3-D. We extend this work to the case where a constraint on the possible color of the illuminant is factored into the gamut mapping algorithm. We show here that the illumination constraint can be enforced during selection without explicitly intersecting the two constraint sets. In the final part of this paper we reappraise the selection task. Gamut mapping returns the set of feasible illuminant maps. Our new algorithm is tested using real and synthetic images. The results of these tests show that the algorithm presented delivers excellent color constancy.     

Psychophysical evaluation of gamut mapping techniques using simplerendered images and artificial gamut boundaries

Using a paired comparison paradigm, various gamut mapping algorithms were evaluated using simple rendered images and artificial gamut boundaries. The test images consisted of simple rendered spheres floating in front of a gray background. Using CIELAB as our device-independent color space, cut-off values for lightness and chroma, based on the statistics of the images, were chosen to reduce the gamuts for the test images. The gamut mapping algorithms consisted of combinations of clipping and mapping the original gamut in linear piecewise segments. Complete color space compression in RGB and CIELAB was also tested. Each of the colored originals (R,G,B,C,M,Y, and Skin) were mapped separately in lightness and chroma. In addition, each algorithm was implemented with saturation (C(*)/L(*)) allowed to vary or retain the same values as in the original image. Pairs of test images with reduced color gamuts were presented to twenty subjects along with the original image. For each pair the subjects chose the test image that better reproduced the original. Rank orders and interval scales of algorithm performance with confidence limits were then derived. Clipping all out-of-gamut colors was the best method for mapping chroma. For lightness mapping at low lightness levels and high lightness levels particular gamut mapping algorithms consistently produced images chosen as most like the original. The choice of device-independent color space may also influence which gamut mapping algorithms are best.     

语义驱动的颜色恒常决策算法

不同颜色恒常性算法适用于不同场景下的图像,算法融合是扩展颜色恒常性算法适用范围常用的方法之一,而现有融合性算法在算法选择依据上忽略了语义信息在图像纹理特征描述中的作用,导致光源估计时的精度不高。针对该问题,提出一种语义驱动的颜色恒常决策算法。首先,利用PSPNet(Pyramid Scene Parsing Network)模型对经过一阶灰度边缘算法(1st Gray Edge)偏色预处理后的目标图像进行场景语义分割,并计算场景中各个语义类别的占比;其次,根据语义类别及占比在已训练的决策集合中寻找相似的参考图像,并使用欧氏距离计算两者的语义相似度;最后,将语义相似度与基于多维欧氏空间确定的阈值进行判别,根据判别结果选择合适算法为目标图像实行偏色校正。在Color Checker和NUS-8 camera两种数据集中的实验结果表明,所提算法光源估计角度误差较单一算法均大幅度下降,且较同类型融合性算法分别下降14.02%和8.17%,提高了光源估计的鲁棒性和准确度。     

Color image enhancement with high saturation using piecewise linear gamut mapping

Most of the color image enhancement algorithms are implemented in two stages: gray scale image enhancement, which finds the target intensity, and then gamut mapping of the original color coordinates to the target. Therefore, hue preserving gamut mapping is an essential and crucial step, which influences colorfulness. In conventional color mapping methods, color saturation is reduced after intensity modification, which deteriorates subjective image quality. In this paper, a new color enhancement algorithm resulting in high color saturation is proposed. The proposed method employs multiplicative and additive color mapping to improve color saturation without clipping of a color component for increased target intensity as well as decreased cases. This new scheme is fast and effective, therefore, it can be employed to real time applications such as video signal processing.     

Using Zernike moments for the illumination and geometry invariantclassification of multispectral texture

We develop a method for recognizing color texture independent of rotation, scale, and illumination. Color texture is modeled using spatial correlation functions defined within and between sensor bands. Using a linear model for surface spectral reflectance with the same number of parameters as the number of sensor classes, we show that illumination and geometry changes in the scene correspond to a linear transformation of the correlation functions and a linear transformation of their coordinates. A several step algorithm that includes scale estimation and correlation moment computation is used to achieve the invariance. The key to the method is the new result that illumination, rotation, and scale changes in the scene correspond to a specific transformation of correlation function Zernike moment matrices. These matrices can be estimated from a color image. This relationship is used to derive an efficient algorithm for recognition. The algorithm is substantiated using classification results on over 200 images of color textures obtained under various illumination conditions and geometric configurations.     

Black light: how sensors filter spectral variation of theilluminant

Visual sensor responses may be used to classify objects on the basis of their surface reflectance functions. In a color image, the image data are represented as a vector of sensor responses at each point in the image. This vector depends both on the surface reflectance function and on the spectral power distribution of the ambient illumination. Algorithms designed to classify objects on the basis of their surface reflectance functions typically attempt to overcome the dependence of the sensor responses on the illuminant by integrating sensor data collected from multiple surfaces. In machine vision applications, we show that it is often possible to design the sensor spectral responsivities so that the vector direction of the sensor responses does not depend upon the illuminant. We state the conditions under which this is possible and perform an illustrative calculation. In biological systems, where the sensor responsivities are fixed, we show that some changes in the illumination cause no change in the sensor responses. We call such changes in illuminant black illuminants. It is possible to express any illuminant as the sum of two unique components. One component is a black illuminant. We call the second component the visible component. The visible component of an illuminant completely characterizes the effect of the illuminant on the vector of sensor responses.     

Color constancy by category correlation

Finding color representations that are stable to illuminant changes is still an open problem in computer vision. Until now, most approaches have been based on physical constraints or statistical assumptions derived from the scene, whereas very little attention has been paid to the effects that selected illuminants have on the final color image representation. The novelty of this paper is to propose perceptual constraints that are computed on the corrected images. We define the category hypothesis, which weights the set of feasible illuminants according to their ability to map the corrected image onto specific colors. Here, we choose these colors as the universal color categories related to basic linguistic terms, which have been psychophysically measured. These color categories encode natural color statistics, and their relevance across different cultures is indicated by the fact that they have received a common color name. From this category hypothesis, we propose a fast implementation that allows the sampling of a large set of illuminants. Experiments prove that our method rivals current state-of-art performance without the need for training algorithmic parameters. Additionally, the method can be used as a framework to insert top-down information from other sources, thus opening further research directions in solving for color constancy.     

Gamut Compression and Extension Algorithms Based on Observer Experimental Data

Gamut compression algorithms have traditionally been defined functionally and then tested with deductive methods, e.g., psychophysical experiments. Our study offers an alternative, an inductive method, in which observers judge image colors to represent the original images more accurately. We developed a computer‐controlled interactive tool that modifies the color appearance of pictorial images displayed on a monitor. In experiments, observers used the tool to alter color pixels according to the region of color space to which they belonged. We created three different gamut compression algorithms based on the observer experimental data. Observer groups evaluated the performance of the newly‐developed algorithms, existing gamut compression algorithms, and an image based on the average observers’ results from experiments in this study. The study of gamut extension is unlike the study of gamut compression in that it mainly deals with the degree of image pleasantness as judged by observers. The results of the gamut extension experiments in this study not only make available worthwhile data but also suggest a methodology for using the observer experimental tool for future gamut extension research.     

Recovering colors in an image with chromatic illuminant

Colors in a scene change under different illuminants. By adopting models that are used to describe human color constancy, the maximum-spectral-value method is proposed to estimate the illuminant from the maximum distribution of reflected lights in an image. From the experimental results, the proposed method recovers colors well with different illuminants.     

Geometrically resilient color image zero-watermarking algorithm based on quaternion Exponent moments

Although research on zero-watermarking has made great progress in recent years, most of it has been focused on grayscale images rather than color ones, and cannot resist geometric attacks efficiently. In this paper, we discuss properties of quaternion Exponent moments (QEMs) in detail and propose a robust color image zero-watermarking algorithm which is robust to geometric attacks. We first compute and select robust QEMs of the original color image, and then a binary feature image is constructed using the magnitude of the selected moments. Eventually, a bitwise exclusive-or is applied on the binary feature image and a scrambled binary logo to generate the zero-watermark image. Experimental results show that the proposed zero-watermarking algorithm is robust to both geometric attacks and common image processing attacks effectively. Compared to similar zero-watermarking algorithms and traditional watermarking algorithms based on QEMs, the proposed zero-watermarking algorithm has better performance.     

A multilinear constraint on dichromatic planes for illumination estimation.

A new multilinear constraint on the color of the scene illuminant based on the dichromatic reflection model is proposed. The formulation avoids the problem, common to previous dichromatic methods, of having to first identify pixels corresponding to the same surface material. Once pixels from two or more materials have been identified, their corresponding dichromatic planes can be intersected to yield the illuminant color. However, it is not always easy to determine which pixels from an arbitrary region of an image belong to which dichromatic plane. The image region may cover an area of the scene encompassing several different materials and, hence, pixels from several different dichromatic planes. The new multilinear constraint accounts for this multiplicity of materials and provides a mechanism for choosing the most plausible illuminant from a finite set of candidate illuminants. The performance of this new method is tested on a database of real images.     

Gamut mapping optimization algorithm based on gamut-mapped image measure (GMIM)

This paper presented a novel gamut mapping optimization algorithm based on a special designed gamut-mapped image measure (GMIM). We formulated GMIM as an objective function in an iterative manner, where the iterative step changes based on the information weight which can be computed between the original image and the initial gamut-mapped images. A new metric, GMIM, is developed to compute the achromatic differences and the chromatic differences of the input images based on structural similarity index. For the achromatic part, we introduce the gradient information to detect the achromatic distortions between two images; the chromatic part aims to analyze the hue and chroma information. Further, the circular statistical theory is employed to calculate the hue value. In the iterative process, we change the iterative steps according to the information weight which can be computed by the information theory. The information map between images indicated regions in an image which human paid attentions to. Experimental results demonstrated that our new gamut mapping method can preserve the brightness, color, as well as detail information of the reference images.     

Retaining local image information in gamut mapping algorithms.

Our topic is the potential of combining global gamut mapping with spatial methods to retain the percepted local image information in gamut mapping algorithms. The main goal is to recover the original local contrast between neighboring pixels in addition to the usual optimization of preserving lightness, saturation, and global contrast. Special emphasis is placed on avoiding artifacts introduced by the gamut mapping algorithm itself. We present an unsharp masking technique based on an edge-preserving smoothing algorithm allowing to avoid halo artifacts. The good performance of the presented approach is verified by a psycho-visual experiment using newspaper printing as a representative of a small destination gamut application. Furthermore, the improved mapping properties are documented with local mapping histograms.     

A comparison of computational color constancy Algorithms. II. Experiments with image data

For pt.I see ibid., vol. 11, no.9, p.972-84 (2002). We test a number of the leading computational color constancy algorithms using a comprehensive set of images. These were of 33 different scenes under 11 different sources representative of common illumination conditions. The algorithms studied include two gray world methods, a version of the Retinex method, several variants of Forsyth's (1990) gamut-mapping method, Cardei et al.'s (2000) neural net method, and Finlayson et al.'s color by correlation method (Finlayson et al. 1997, 2001; Hubel and Finlayson 2000). We discuss a number of issues in applying color constancy ideas to image data, and study in depth the effect of different preprocessing strategies. We compare the performance of the algorithms on image data with their performance on synthesized data. All data used for this study are available online at http://www.cs.sfu.ca//spl sim/color/data, and implementations for most of the algorithms are also available (http://www.cs.sfu.ca//spl sim/color/code). Experiments with synthesized data (part one of this paper) suggested that the methods which emphasize the use of the input data statistics, specifically color by correlation and the neural net algorithm, are potentially the most effective at estimating the chromaticity of the scene illuminant. Unfortunately, we were unable to realize comparable performance on real images. Here exploiting pixel intensity proved to be more beneficial than exploiting the details of image chromaticity statistics, and the three-dimensional (3-D) gamut-mapping algorithms gave the best performance.     

多特征融合的自适应加权采样上下文感知相关滤波跟踪算法

针对传统相关滤波器使用特征单一、背景信息不足等缺点,提出一种多特征融合的自适应加权采样的上下文感知相关滤波算法.首先,对于灰度图像序列,采用方向梯度直方图(FHOG)特征、局部二值模式(LBP)特征以及灰度特征相融合;对于彩色图像序列,则采用方向梯度直方图(FHOG)特征、局部二值模式(LBP)特征以及颜色(CN)特征...     

一种基于Retinex理论的图像增强算法

提出了一种基于Retinex理论的图像增强算法。Retinex理论的实质就是从图像S中抛开照射光L的影响来获得物体的反射性质R,即获得物体本来的面貌。在对Retinex理论进行研究的基础上,论述了该算法的原理和实现方法,并通过实验与几种传统的图像增强方法以及单尺度、多尺度和带彩色恢复的多尺度Retinex进行了比较,表明该算法对于图像阴影细节的增强和色彩的保真较一般的图像增强算法能够达到更好的效果。     

A comparison of computational color constancy algorithms. I: Methodology and experiments with synthesized data

We introduce a context for testing computational color constancy, specify our approach to the implementation of a number of the leading algorithms, and report the results of three experiments using synthesized data. Experiments using synthesized data are important because the ground truth is known, possible confounds due to camera characterization and pre-processing are absent, and various factors affecting color constancy can be efficiently investigated because they can be manipulated individually and precisely. The algorithms chosen for close study include two gray world methods, a limiting case of a version of the Retinex method, a number of variants of Forsyth's (1990) gamut-mapping method, Cardei et al.'s (2000) neural net method, and Finlayson et al.'s color by correlation method (Finlayson et al. 1997, 2001; Hubel and Finlayson 2000) . We investigate the ability of these algorithms to make estimates of three different color constancy quantities: the chromaticity of the scene illuminant, the overall magnitude of that illuminant, and a corrected, illumination invariant, image. We consider algorithm performance as a function of the number of surfaces in scenes generated from reflectance spectra, the relative effect on the algorithms of added specularities, and the effect of subsequent clipping of the data. All data is available on-line at http://www.cs.sfu.ca//spl sim/color/data, and implementations for most of the algorithms are also available (http://www.cs.sfu.ca//spl sim/color/code).     

Representations of Relative Display Gamut Size

Gamut size of a wide gamut display is usually represented with the ratio of its chromaticity triangle area and the chromaticity triangle area of a standard display in CIE xy color coordinates. Such a chromaticity area ratio (CAR) is a rough relative gamut size, because CIE xy is a nonuniform color coordinate system and display gamut is a volume in color space. The representation of relative gamut size with the ratio of the discernible color numbers in a display gamut and NTSC TV gamut is studied. This ratio is called the discernible color number ratio (DCNR). Discernible color number is counted with CIE94 color difference formula in CIELAB color space. It is found that CAR is larger than DCNR for the display with primary purity higher than NTSC primary purity. For example, the CAR of a practical light-emitting diode (LED) display with respect to NTSC TV is 7.8% overestimated, in which red, green, and blue LED primary wavelengths are 625, 520, and 470 nm, respectively; red, green, and blue LED bandwidths are 20, 40, and 30 nm, respectively. In addition, the DCNRs of wide gamut displays with respect to the object color of the same white illuminant are investigated. It is shown that the gamut size improvement for laser display compared with LED display is not significant.     

A multiscale framework for spatial gamut mapping.

Image reproduction devices, such as displays or printers, can reproduce only a limited set of colors, denoted the color gamut. The gamut depends on both theoretical and technical limitations. Reproduction device gamuts are significantly different from acquisition device gamuts. These facts raise the problem of reproducing similar color images across different devices. This is well known as the gamut mapping problem. Gamut mapping algorithms have been developed mainly using colorimetric pixel-wise principles, without considering the spatial properties of the image. The recently proposed multilevel gamut mapping approach takes spatial properties into account and has been demonstrated to outperform spatially invariant approaches. However, they have some important drawbacks. To analyze these drawbacks, we build a common framework that encompasses at least two important previous multilevel gamut mapping algorithms. Then, when the causes of the drawbacks are understood, we solve the typical problem of possible hue shifts. Next, we design appropriate operators and functions to strongly reduce both haloing and possible undesired over compression. We use challenging synthetic images, as well as real photographs, to practically show that the improvements give the expected results.