A novel image compression method for medical images using geometrical regularity of image structure

Abstract

This paper proposes a novel medical image compression method using sparse representation approach that exploits the geometrical regularity of image structure. The geometric flow represents the direction in which the image gray levels have regular variations. The wavelet decomposition of geometric regularized data results in less number of significant coefficients. The directions of regularity are represented using two-dimensional vector field, and the approximation of these directions is obtained using spline representation. The directional decomposition of the image along with geometric flow is further improved by bandelet basis. The bandelet coefficients are encoded using Set Partitioning in Hierarchical Trees encoder, followed by global thresholding with fixed encoding. Experimental results demonstrate that the proposed method provides significant improvement in compression performance over state-of-the-art compression methods.