Block truncation coding using visual patterns for image compression

Block Truncation Coding (BTC) is a block-based image compression technique developed by E. J. Delp and O. R. Mitchell. It has advantages of low computational load, little memory requirement and can achieve 2 bits per pixel (bpp). The aim of this research is to develop new image compression techniques based on the standard BTC algorithm but using less bit rate than BTC. The proposed method encodes images at low bitrate and still with high subjective as well as high objective image quality. But my focus is on the subjective quality. For monochrome images, a few patterns have been designed to represent image features such as uniform region, horizontal, vertical, and diagonal edge regions. These patterns are visually significant to the human observer (so called Visual Patterns) and are designed based on the characteristics of Human Vision System (HVS). Experimential results show that using only a few Visual Patterns achieve high image quality. For color images, several schemes have been proposed for different signal formats (RGB, YIQ, and YUV signals). Hierarchical Clustering (HC), Weighted Trichromatic Transform (WTT) and Downsampling (DS) are techniques employed