Predictive Coding Based Lossless Image Compression Model

Abstract

Image compression is a procedural technique wherein images undergo compression by means of encoding image data using fewer bits and eliminating redundant information. Prior to storage or transmission via a communication medium or network, image data undergoes compression. Various methods are employed for the encoding and decoding of images within the digital image compression process. Due to the critical nature of the medical and military image data, most obvious choice is a lossless compression technique. The predictive coding method consists of a prediction, context modeling and entropy coding.  Since, current predictors don’t work efficiently for different types of images due to its performance limitation in different portions of the image. To overcome this problem, the research target is to develop a context-sensitive method by making clustering of input image. And calculate context of pixels by calculating 20-Dimensional difference vector. Update the weights of clusters after calculating prediction errors of pixel. Finally, by applying Golomb coding after updating of cluster’s weights, output stream image is compressed with fewer bits. Proposed method obtains a relatively higher compression ratio of continuous-tone images than other lossless predictive coding techniques. Proposed method reduces the compressed stream size by 13%. This is a significant improvement to store images with fewer bits after compression.