Image Compression and Edges Extraction using Fast Transforms by Selecting an Appropriate Region Samples Shape

Abstract

This research is a study of region-based image compression techniques using discrete cosine transform (DCT) and fast Walsh transform (FWT). The process of image compression goes through several stages, the most important conversion of the digital image to the gray level. The other stage is the selection of the zone-of-interest region to perform the compression process which is known as zonal compression. These existed methods are selected according to the six different geometric shapes that are proposed in this work to be compared with the existing methods. This study was also used to illustrate how to extract the image edges using low-pass and high-pass (LPF and HPF) filters in order to distinguish the details of the digital image and to clarify its boundaries using the Threshold technique on the image obtained by HPF. This research aims at compression the digital image in order to reduce its size without affecting the details of the image and clarify its boundaries. The comparison is done between (DCT) and (FWT) transforms in terms of peak-to-signal ratio (PSNR), bit per pixel (bpp), mean square error (MSE), and compression ratio (CR) and then we apply the threshold technology to the resulting image from (HPF) to obtain the binary image that illustrates the image boundary. The resulting image is called the image detailed edges. Experimental results and analysis for image compression using LPF show that DCT gives higher PSNR (45 decibels) with CR=89.0144%than FWT (36 decibels) with CR=79.6921%; while the FWT introduce better quality than that by using DCT for contour extraction using HPF.