A Novel Method of Lossless Compression of Images in the Necessary Region of Interest using Golomb Coding

This project is developed to overcome the problem of compression in hyperspectral images. We are concentrating on necessary region of interest (ROI) in hyperspectral images and doing lossless compression on these regions alone than the no-data regions. For this, we use a two-stage prediction scheme, context –similarity based weighted average filtering to remove redundancy for ROI of 2-D spatial image followed by recursive least square filtering to decor relate the hyperspectral images for compression. Then, Golomb- Rice Code is applied for the residuals of full-context pixels and boundary pixels of earlier stage. We use code book to provide better quality, compression and de-compression and also they serve as reference for color and shape of the object. The coding gains of the GR code is studied using mixture geometric model to represent the residuals associated with the pixels.