Hindawi Publishing Corporation EURASIP Journal on Image and Video Processing Volume 2010, Article ID 837237, 22 pages doi:10.1155/2010/837237 Research Article Perceptually Motivated Automatic Color Contrast Enhancement Based on Color Constancy Estimation Anustup Choudhury and G´ erard Medioni Department of Computer Science, University of Southern California, Los Angeles, CA 90089, USA Correspondence should be addressed to G´ erard Medioni, medioni@usc.edu Received 30 March 2010; Revised 20 August 2010; Accepted 1 October 2010 Academic Editor: Zhou Wang Copyright © 2010 A. Choudhury and G. Medioni. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. We address the problem of contrast enhancement for color images. Our method to enhance images is inspired from the retinex theory. We try to estimate the illumination and separate it from the reflectance component of an image. We use denoising techniques to estimate the illumination and while doing so achieve color constancy. We enhance only the illumination component of the image. The parameters for enhancement are estimated automatically. This enhanced illumination is then multiplied with the reflectance to obtain enhanced images with better contrast. We provide validation of our color constancy approach and show performance better than state-of-the-art approaches. We also show “visually better” results while comparing our enhancement results with those from other enhancement techniques and from commercial software packages. We perform statistical analysis of our results and quantitatively show that our approach produces effective image enhancement. This is validated by ratings from human observers. 1. Introduction The human visual system (HVS) is a sophisticated mecha- nism capable of capturing a scene with very precise repre- sentation of detail and color. In the HVS, while individual sensing elements can only distinguish limited quantized levels, the entire system handles large dynamic range through various biological actions. Current capture or display devices cannot faithfully represent the entire dynamic range of the scene, therefore images taken from a camera or displayed on monitors/display devices suffer from certain limitations. As a result, bright regions of the image may appear overexposed and dark regions may appear underexposed. The objective of contrast enhancement is to improve the visual quality of images. One of the most common techniques to enhance the contrast of images is to perform histogram equalization. The advantage of this technique is that it works very well for grayscale images; however, when histogram equalization is used to enhance color images, it may cause a shift in the color scale, resulting in artifacts and an imbalance in image color as shown in Figure 1. These unwanted artifacts are not desirable, as it is critical to maintain the color properties of the image while enhancing them. A high-level overview of our approach can be seen in Figure 2. We assume any image to be a pixel-by-pixel product of the illumination (light that falls on the scene) and the reflectance component of the scene. This can be expressed as I(x) = L(x)R(x), (1) where L(x) is the illumination component and R(x) is the reflectance component of the image I(x) with spatial coordinate x. In this paper, we deal with color images. So, I(x), L(x), and R(x) have 3 components—one for each color channel. For instance, for the illumination image, L(x), we denote the red color channel by L red (x), the green channel by L green (x), and the blue color channel by L blue (x). Similarly, we can denote the color channels for other images. The capital bold font denotes multiband. The presence of a bar above the capital bold font denotes singleband. Similarly, the spatial coordinate, x = (x, y) ∈ R 2 .