<title>New quadtree predictive image coding technique using pattern-based classification</title>

Abstract

A new quadtree segmented predictive image coding technique is presented in this paper for exploiting the correlation between adjacent image blocks and uniformity in variable block size image blocks. For exploiting correlation between adjacent image blocks, a predictive coding technique is used for reducing the inter block correlation. The proposed quadtree technique decomposes an image into variable size block, and each segmented block is then coded at a different bit rate according to the level of visual activity inside the block. A novel classification scheme, which operates based on the distribution of the block residuals, is employed to determine the activity level inside the block. In this method, the orientation of the pattern appearing inside the block will be computed as an aid to the classification. To preserve edge integrity, a block pattern based coding technique is proposed and incorporated to the predictive coding method for coding the high-activity blocks of the segmented image. The use of a set of parameters associated with the pattern appearing inside a high activity block at the receiver, together with the inter block correlation information, reduce the cost of instruction and saves the encoding time. Experiments have been conducted to compare with other predictive and quadtree- based techniques. Results show a lower bit rate at competitive reconstruction quality.© (2003) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.