PC-based image compression using adaptive rectangular segmentation

Abstract

A method of image compression, along with its suitability for PC-implementation, is considered. The compressed file is a description of a set of rectangular segments of the image whose pixels all have approximately the same intensity. The compression process developed in this work was implemented on a PC 33 MHz -386 machine using Turbo C-2.0. The process produces a binary tree of rectangular segments whose pixel intensity variance is below a given amount. It does this by recursively dividing rectangular image segments into two, preferably along lines of high intensity gradient. The divisions are found by trying every possible horizontal and vertical cut for the segment. The process is accelerated by two preprocessed moment tables that replace much of the repetitive calculating. The resulting segments are compactly described in the compressed file. The decompression routine recreates the same binary tree to place and fill the segments.