Robust Color Correction for Preserving Spatial Variations within Photographs

Abstract

Figure 1: A reference colorchart (left image) is commonly used for color correction which is an ill-posed problem. State-of-the-art root-polynomial regressionmethod reducesCIE XYZ or linear-RGB color differences for the transformed reference blocks in themean-sense. It improves significantly with the increasing regression order as CIEΔE is seen to drop with the increasing order. However, it does not account for spatial variations and produces serious artifacts as demonstrated here (center image). Proposed method improves color correction while preserving spatial variations, white-balancing appropriately and not over-damping the luminance as reported by Varghese et al. for their CIE_ΔE minimizing method.