Image Color Reduction Using Progressive Histogram Quantization and Kmeans Clustering

Abstract

Color reduction is an important tool for different image processing and computer vision applications. In this paper, a progressive histogram-based color reduction (quantization) is used with the Kmeans clustering algorithm to speed up the quantization process of the Kmeans method. The progressive histogram quantization (PHQ) is a simple iterative algorithm where a single histogram bin is merged to one of its two nearest neighbors’ bins at each iteration. The histogram bin is merged according to the differences in the value (pixel counts) and the location of the left and right bins. The PHQ algorithm is used as a pre-quantization for the Kmeans clustering to reduce the size of the data and speed up the clustering process. The experimental results show that the PHQ+Kmeans algorithm maintains good image quality and enhances the execution time compared to the Kmeans clustering algorithm alone when applied on remote sensing images.