A Color to Grayscale Conversion Based on Singular Value Decomposition and Extended Bimodal Gaussian Models

Aiming at the problems of insufficient contrast, blurred details and lack of hierarchy in traditional decolorization methods for color images, a new contrast preserving decolorization method is proposed in this paper by combining the singular value decomposition (SVD) and the improved bimodal Gaussian model. Firstly, the idea of SVD dimensionality reduction is introduced into the decolorization process of color images, so that the produced grayscale image can better maintain the sense of hierarchy and global visual contrast of the initial color image. Secondly, the range of weight parameters in the bimodal Gaussian function model is extended from [0,1] to [-1,1], in order to have more selectable weight values when solving the bimodal Gaussian model by discrete search method, so that the produced grayscale image can better maintain the local visual contrast and detailed features of the color initial image. Finally, an efficient discrete search method is used to quickly solve the proposed decolorization model. The simulation experimental results show that, compared with the traditional decolorization methods, the new method proposed in this paper can better maintain the contrast, hierarchy and details of the color initial image, and outperforms the other classical decolorization methods in terms of both subjective and objective evaluations.