Color Characterization and Balancing by a Nonlinear Line Attractor Network for Image Enhancement

Abstract

A novel method to map high dynamic range scenes to low dynamic range images utilizing the concept of color characterization, enhancement, and balancing is described in this letter. Each pixel of the image is first characterized by extracting the relationship of the red, green, and blue components along with its corresponding neighbors using a nonlinear line attractor network to form an associative memory. Then, the illumination enhancement process is performed using a hyperbolic tangent function to provide dynamic range compression to each pixel in the image. The slope of the hyperbolic tangent function is controlled using a parameter that is determined by the local and global statistics of the image to facilitate the change of the intensity level. A color balancing process restores the original color characteristics of the image based on learned associative memory matrices which eliminate image distortion due to improper recombination of red, green and blue components after enhancement. Experiments conducted on images captured at extremely uneven lighting environments show that the proposed method outperforms other image enhancement algorithms.