A new fast color reduction method based on adaptive histogram binning approach

Abstract

- Most color reduction methods that are based on image clustering in a 3D color space have extremely high computational costs, especially for large size images. In this paper, a new fast adaptive color reduction method is proposed which, computationally, is independent of the image size and reduces the pixel depth from 24 bits (used to represent tristimulus values in the most commonly hardware-oriented RGB model) to a maximum of 15 bits. To achieve this purpose, by introducing a new hybrid cost function and using a modified version of the Gravitational Search Algorithm (GSA), an adaptive histogram binning approach has been developed. Although the cube re-quantization accuracy in the histogram binning approach is lower compared to the 3D data clustering method, it leads to a significant reduction in computational cost. In this paper, while taking this advantage, we seek to reduce re-quantization error using the adaptive histogram binning of RGB color components. Despite a significant reduction in pixel depth, the proposed color reduction approach, due to the adaptive reduction of image colors, results in an appropriate color reduction for a wide variety of images.