MODEL BASED IMAGE COMPRESSION FOR TELEMEDICINE

Abstract

Image compression plays a key role in medical image archiving and transfer. Current compression schemes produce high compression rates, if loss of quality is acceptable. However, deficiency in quality cannot be accepted in the field of medicine. Any loss in information may lead to wrong decisions. Therefore a compression scheme is required which provides high compression ratio and is lossless. The scheme proposed in this dissertation work. is lossless and provides high compression ratio. The proposed approach is model based and requires two major operations: registration of model. and input images, and compression of residual image. A novel Quadtree based Adaptive Block Partitioning with Rearrangement (QABPR) compression scheme has been proposed. The idea of using quad-tree as a primary data structure came after seeing the effective usage of quad-trees in fractal based compression algorithms. The idea of adaptive block partitioning is taken from various segmentation algorithms. Rearrangement was done to exploit the redundancies in blocks with similar intensities. Results have shown that proposed compression scheme works better than other compression schemes such as Huffman coding, GIF, JPEG-LS and JPEG2000. Registration is done using minimization of RIS (Residual Image Size). Main idea of the proposed registration scheme is to use a criterion for alignment, which leads to maximum compression ratio. Results shows that proposed registration improves