Effective Hybrid Compression Model for Medical Images

The main objective of this work is to propose efficient ROI based hybrid compression models that can efficiently extract the ROI by segmentation and compress medical images very effectively with good level of visual quality. The work proposes a multiple approach of extracting the ROI such as the sequence of morphological operations for MR Brain images gradient based approaches and morphological operations for CT Abdomen images and ultra-contour models & structural edge detectors for CT Lung images.To improve the performance of the compression the Convolutional Neural Network (CNN) based segmentation method is used for ROI extraction of MR brain images and the extracted region is compressed with BPT (Binary Plane Technique) operated in both lossy and lossless for NROI and ROI. It is found that the DNN (Deep Neural Network) approach is attained better segmentation efficiency when compared with Satheesh’s approach in terms of accuracy similarity index and correct detection ratio. The efficiency is improved by 6% than earlier methods.