Abstract

Color-to-Gray scale conversion methods try to identify weights for various color channels to obtain a gray-scale image. These weights can be fixed either globally or computed on a localized basis. This paper presents an approach for computing the global weights using localized regions perpetually selected based on human perception. The approach aims to bring forth a color invariant gray scale conversion, such that it tries to maximize the required foreground information. The proposed method was tested on DIBCO-2013 dataset and qualitatively evaluated by looking at the structural similarity with the foreground using SSIM. The experimental results of ours and other color-to-gray scale methods have been tabulated and discussed.