Fusion model of Modified Wavelet Transform and Neural Network for medical image compression

Abstract

Medical imaging is one of emerging fields that has high impact on analysis and diagnosis of diseases. Image compression is introduced in the medical image processing to solve memory and bandwidth requirement problems. In this paper, a hybrid image compression method by integrating Modified Wavelet Transform (MWT) and Neural Network (NN) is proposed for efficient storage as well as transmission of medical images. Primarily, medical images are preprocessed using median filter to suppress noisy data and then decomposed into approximation and detailed coefficients using wavelet transform. Subsequently, detailed coefficients are filtered with proposed thresholding technique. Both approximation and detailed coefficients are divided into sub matrix with the size of n x n and used as input to the NN to perform compression. The image compression performance of the proposed method is evaluated by computing peak signal to noise ratio, compression ratio and storage space and compared with prevailing methods to prove its effectiveness. The empirical findings reveal that the proposed method provides best performance compared to the former methods taken for comparison.