Color names are required in real-world applications such as image retrieval and image annotation. Traditionally, they are learned from a collection of labeled color chips. These color chips are labeled with color names within a well-defined experimental setup by human test subjects. However, naming colors in real-world images differs significantly from this experimental setting. In this paper, we investigate how color names learned from color chips compare to color names learned from real-world images. To avoid hand labeling real-world images with color names, we use Google image to collect a data set. Due to the limitations of Google image, this data set contains a substantial quantity of wrongly labeled data. We propose several variants of the PLSA model to learn color names from this noisy data. Experimental results show that color names learned from real-world images significantly outperform color names learned from labeled color chips for both image retrieval and image annotation.

/pdf/learning-color-names-for-real-world-applications-4br6siq4al.pdf

Learning Color Names for Real-World Applications

Computational color constancy is a fundamental prerequisite for many computer vision applications. This paper presents a survey of many recent developments and state-of-the-art methods. Several criteria are proposed that are used to assess the approaches. A taxonomy of existing algorithms is proposed and methods are separated in three groups: static methods, gamut-based methods, and learning-based methods. Further, the experimental setup is discussed including an overview of publicly available datasets. Finally, various freely available methods, of which some are considered to be state of the art, are evaluated on two datasets.

Computational Color Constancy: Survey and Experiments

We want to integrate colourfulness in an image quality evaluation framework. This quality framework is meant to evaluate the perceptual impact of a compression algorithm or an error prone communication channel on the quality of an image. The image might go through various enhancement or compression algorithms, resulting in a different -- but not necessarily worse -- image. In other words, we will measure quality but not fidelity to the original picture.While modern colour appearance models are able to predict the perception of colourfulness of simple patches on uniform backgrounds, there is no agreement on how to measure the overall colourfulness of a picture of a natural scene. We try to quantify the colourfulness in natural images to perceptually qualify the effect that processing or coding has on colour. We set up a psychophysical category scaling experiment, and ask people to rate images using 7 categories of colourfulness. We then fit a metric to the results, and obtain a correlation of over 90% with the experimental data. The metric is meant to be used real time on video streams. We ignored any issues related to hue in this paper.

/pdf/measuring-colorfulness-in-natural-images-2mpesoy4lg.pdf

Measuring colorfulness in natural images

Color science is a multidisciplinary field with broad applications in industries such as digital imaging, coatings and textiles, food, lighting, archiving, art, and fashion. Accurate definition and measurement of color appearance is a challenging task that directly affects color reproduction in such applications. Color Appearance Models addresses those challenges and offers insight into the preferred solutions. Extensive research on the human visual system (HVS) and color vision has been performed in the last century, and this book contains a good overview of the most important and relevant literature regarding color appearance models.

https://www.spiedigitallibrary.org/journals/Journal-of-Electronic-Imaging/volume-23/issue-2/029901/Color-Appearance-Models/10.1117/1.JEI.23.2.029901.pdf

Color Appearance Models

This landmark book is the first to describe HDRI technology in its entirety and covers a wide-range of topics, from capture devices to tone reproduction and image-based lighting. The techniques described enable you to produce images that have a dynamic range much closer to that found in the real world, leading to an unparalleled visual experience. As both an introduction to the field and an authoritative technical reference, it is essential to anyone working with images, whether in computer graphics, film, video, photography, or lighting design.



New material includes chapters on High Dynamic Range Video Encoding, High Dynamic Range Image Encoding, and High Dynammic Range Display Devices
Written by the inventors and initial implementors of High Dynamic Range Imaging
Covers the basic concepts (including just enough about human vision to explain why HDR images are necessary), image capture, image encoding, file formats, display techniques, tone mapping for lower dynamic range display, and the use of HDR images and calculations in 3D rendering
Range and depth of coverage is good for the knowledgeable researcher as well as those who are just starting to learn about High Dynamic Range imaging


Table of Contents

Introduction; Light and Color; HDR Image Encodings; HDR Video Encodings; HDR Image and Video Capture; Display Devices; The Human Visual System and HDR Tone Mapping; Spatial Tone Reproduction; Frequency Domain and Gradient Domain Tone Reproduction; Inverse Tone Reproduction; Visible Difference Predictors; Image-Based Lighting.

High Dynamic Range Imaging: Acquisition, Display, and Image-Based Lighting (The Morgan Kaufmann Series in Computer Graphics)

The CIE Technical Committee 8-01, color appearance models for color management applications, has recently proposed a single set of revisions to the CIECAM97s color appearance model This new model, called CIECAM02, is based on CIECAM97s but includes many revisions1-4 and some simplifications A partial list of revisions includes a linear chromatic adaptation transform, a new non-linear response compression function and modifications to the calculations for the perceptual attribute correlates The format of this paper is an annotated description of the forward equations for the model Introduction The CIECAM02 color appearance model builds upon the basic structure and form of the CIECAM97s5,6 color appearance model This document describes the single set of revisions to the CIECAM97s model that make up the CIECAM02 color appearance model There were many, often conflicting, considerations such as compatibility with CIECAM97s, prediction performance, computational complexity, invertibility and other factors The format for this paper will differ from previous papers introducing a color appearance model Often a general description of the model is provided, then discussion about its performance and finally the forward and inverse equations are listed separately in an appendix Performance of the CIECAM02 model will be described elsewhere7 and for the purposes of brevity this paper will focus on the forward model Specifically, this paper will attempt to document the decisions that went into the design of CIECAM02 For a complete description of the forward and inverse equations, as well as usage guidelines, interested readers are urged to refer to the TC 8-01 web site8 or to the CIE for the latest draft or final copy of the technical report This paper is not intended to provide a definitive reference for implementing CIECAM02 but as an introduction to the model and a summary of its structure Data Sets The CIECAM02 model, like CIECAM97s, is based primarily on a set corresponding colors experiments and a collection of color appearance experiments The corresponding color data sets9,10 were used for the optimization of the chromatic adaptation transform and the D factor The LUTCHI color appearance data11,12 was the basis for optimization of the perceptual attribute correlates Other data sets and spaces were also considered The NCS system was a reference for the e and hue fitting The chroma scaling was also compared to the Munsell Book of Color Finally, the saturation equation was based heavily on recent experimental data13 Summary of Forward Model A color appearance model14,15 provides a viewing condition specific means for transforming tristimulus values to or from perceptual attribute correlates The two major pieces of this model are a chromatic adaptation transform and equations for computing correlates of perceptual attributes, such as brightness, lightness, chroma, saturation, colorfulness and hue The chromatic adaptation transform takes into account changes in the chromaticity of the adopted white point In addition, the luminance of the adopted white point can influence the degree to which an observer adapts to that white point The degree of adaptation or D factor is therefore another aspect of the chromatic adaptation transform Generally, between the chromatic adaptation transform and computing perceptual attributes correlates there is also a non-linear response compression The chromatic adaptation transform and D factor was derived based on experimental data from corresponding colors data sets The non-linear response compression was derived based on physiological data and other considerations The perceptual attribute correlates was derived by comparing predictions to magnitude estimation experiments, such as various phases of the LUTCHI data, and other data sets, such as the Munsell Book of Color Finally the entire structure of the model is generally constrained to be invertible in closed form and to take into account a sub-set of color appearance phenomena Viewing Condition Parameters It is convenient to begin by computing viewing condition dependent constants First the surround is selected and then values for F, c and Nc can be read from Table 1 For intermediate surrounds these values can be linearly interpolated2 Table 1 Viewing condition parameters for different surrounds Surround F c Nc Average 10 069 10 Dim 09 059 095 Dark 08 0525 08 The value of FL can be computed using equations 1 and 2, where LA is the luminance of the adapting field in cd/m2 Note that this two piece formula quickly goes to very small values for mesopic and scotopic levels and while it may resemble a cube-root function there are considerable differences between this two-piece function and a cube-root as the luminance of the adapting field gets very small ! k =1/ 5L A +1 ( ) (1) ! F L = 02k 4 5L A ( ) + 01 1" k4 ( ) 2 5L A ( ) 1/ 3 (2) The value n is a function of the luminance factor of the background and provides a very limited model of spatial color appearance The value of n ranges from 0 for a background luminance factor of zero to 1 for a background luminance factor equal to the luminance factor of the adopted white point The n value can then be used to compute Nbb, Ncb and z, which are then used during the computation of several of the perceptual attribute correlates These calculations can be performed once for a given viewing condition

/pdf/the-ciecam02-color-appearance-model-as6tovjk05.pdf

The CIECAM02 color appearance model

Tone reproduction is a key component in image quality and historically tone correction has been performed on a global basis for a given device or image. Several recent investigations have used a variety of algorithms to demonstrate that local color correction provides a significant improvement in image quality relative to a global correction. This paper presents a local color correction operation that is based on non-linear masking. The operation is very fast and does not require any manual adjustments, such as those involved in dodging and burning. The algorithm is roughly equivalent to deriving a specific tone reproduction curve for each pixel in the image. An example image is processed using the proposed masking scheme and the resulting effect is discussed. Finally, the relationship to CIECAM97s is explored and it is proposed that dynamically setting the background parameter results in similar local color corrections. Introduction Tone reproduction is a critical aspect of the perceived quality of a reproduction.

Local Color Correction Using Non-Linear Masking

This paper describes an ongoing web-based approach to collecting color names or color categories. Previous studies have tended to require a large number of observations from a small number of observers. These studies have also tended to limit responses to one-word or monolexical replies. Many studies have also focused on response time or levels of intra-observer agreement in order to identify focal colors. This web-based study uses a distributed design to collect a small number of names from a large number of observers. The responses are neither limited to nor restricted from being monolexical. The focal color analysis is then based on statistical analysis of monolexcially named colors. This paper presents the methodology and infrastructure, as well as considerations for data analysis. Finally, preliminary results of the experiment are results are considered. The data from over 700 participants results in CIELAB hues and lightnesses for the basic colors that agree with previous investigations as well as those investigations agree with each other.

Unconstrained web-based color naming experiment

A method for assigning a lexical classifier to characterize a visual attribute corresponding to an image element forming part of an image. The method involves capturing an initial attribute value for the image element and transforming the captured initial attribute value to a lexical classifier. Transformation involves reference to a database including a set of lexical classifiers corresponding to a particular type of visual attribute. The lexical classifier assigned to the visual attribute is recorded for display or further processing. Transformation of the initial attribute value to the lexical classifier involves application of a machine learning algorithm to the database.

Adaptive lexical classification system

An inkjet printing system uses high-resolution printheads holding different color inks to produce true 600 dpi color printing with a single pass of the printheads. By achieving a smaller drop size in the range of 25-50 picoliters for each printhead having a separate ink color (yellow, cyan, magenta, black), a full range of high-resolution color print-quality modes is possible. In an exemplary embodiment with a carriage scan speed of 25 ips, a two-pass draft mode uses a 7.5 kHz firing frequency, a four-pass normal mode uses a 4 kHz firing frequency, and an eight-pass best mode uses a 2 kHz firing frequency. In all of these print modes, each printhead can provide color inkdrop placement on every pixel row during a single pass.

Nathan Moroney

Papers

The CIECAM02 color appearance model

Local Color Correction Using Non-Linear Masking

Unconstrained web-based color naming experiment

Adaptive lexical classification system

High-resolution inkjet printing using color drop placement on every pixel row during a single pass