IMPLEMENTATION AND EVALUATION OF CONTEXTUAL BASED MEDICAL IMAGE COMPRESSION TECHNIQUES

Abstract

In telemedicine or medical imaging applications very small part of the image will be diagnostically important. So, by compressing region of interest (ROI) in lossless manner and the unimportant background data in a lossy manner, we can save data size and hence transmission time. The medical image compression applications can be broadly divided into two types; (a) image compression for image archival for which the complexity of algorithm & processing time should be less; (b) data browsing / transmission for which progressivity is preferred. In this thesis work, two new methods are proposed for the image archival application. For image transmission applications, ROI coding functionality is added in the JPEG2000 image compression standard named as maxshift method. But, maxshift algorithm is having high complexity and the very less compression ratios. So, a modification is proposed for JPEG 2000 where ROI is transmitted in lossless manner using lossless encoding and background is encoded using JPEG 2000 in progressive manner. SPIHT was designed for optimal progressive transmission. So, if ROI feature is added to SPIHT very high compression ratios can be achieved. So, following methods are proposed based on SPIHT; (a) the ROI is transmitted in lossless manner using lossless encoding and background is encoded using SPIHT named as ESPIHT; (b) scale up (shift) the wavelet coefficients contributing to the ROI reconstruction so that the bits associated with ROI are encoded before background named as MSPIHT; (c) it is based on the observation that any image part occupies the same spatial position in each sub band. So, we can transmit only ROI information in higher level subband (LL) followed by encoding only ROT coefficients and then encoding only background coefficients. As ROI information is transmitted this can be termed as RSPIHT. It is found that wavelet based methods gives better results than discrete cosine transform based method for image archival applications. If ROI selected is of square shape of dimensions having multiples of 2" where n is the level of decomposition, RSPIHT is the best method. If ROI of arbitrary shape is selected, then MSPIHT is the best suited for transmitting medical images over intemet in telemedicine applications.