Adaptive-model predictive lossless compression of medical images

The need for lossless information-preserving medical image data compression has recently increased in order to keep pace with the increasing demands for storage capacity and transmission bandwidth in modern digital medical imaging environments. In this research, a new class of predictive lossless compression techniques is introduced. The class is based on adaptive modification of the prediction model structure to follow the image information source characteristics and is developed to achieve better image decorrelation and, consequently, higher compression rate. Two novel lossless compression techniques that belong to this class; Predictive Classified Lossless Compression and Multi-Model Competitive Lossless Compression that rely on classification and competition, respectively, to adaptively modify the prediction model structure, are developed. The viability of the proposed lossless compression schemes for medical image data is demonstrated by applying them to a large set of medical images of various modalities, resolutions, orientations, and anatomical structures. The techniques are supported by quantitative evaluation of their parameters and comparative studies to different lossless compression techniques in which they have shown a 12-15% improvement in the decorrelated entropy of medical images