Neural Network Based ROI Detection and Hybrid Image Compression

Region of Interest based compression is an efficient method of compression for images with a particular part to be most significant. It is always a better choice to compress the ROI with lossless compression while the rest of image with lossy compression technique. This paper proposes lossless compression for medical image(ROI) and near lossless compression for the rest of the image. Image other than ROI may contain information that is useful, that is why it is appropriate to use near lossless compression for the rest of the image. In this method of compression, PSNR obtained is remarkable and compression ratio can be increased by increasing the base value which is in power of two. Image compression is essential where images need to be stored, transmitted or viewed quickly and efficiently. The artificial neural network is a recent tool in image compression as it processes the data in parallel and hence requires less time and is superior over any other technique. The reason that encourage researchers to use artificial neural networks as an image compression approach are adaptive learning, self organization, noise suppression, fault tolerance and optimized approximations. In this method of compression, medical image will be compressed completely loss less by using morphological operations. In this paper region of interest (ROI) is detected by Artificial Neural Network(ANN) & then compress the part of ROI using lossless method & the part other than ROI compressed by using lossy method. The reason, wavelet method of compression preferred for the rest of the image is, despite of insignificance of region of image other than ROI.