Image Compression Using Deep Learning Based Multi-structure Feature Map and K-Means Clustering

Abstract

Image compression play significant role in the data transfer and storage. Recently, deep learning has achieved tremendous success in various domain of image processing. In this paper, we propose a multi-structure Feature map-based Deep Learning approach with K-means Clustering for image compression. We first use a modified CNN to select a multi-structured region of interest MS-ROI feature map by using several stacked of convolution layers then compress the image by integrating MS-ROI map with K-means. We can establish through experimental results that the proposed approach perform better compared to traditional K-means clustering approach.