A Novel Variable Block-Size Image Compression Based On Edge Detection

Image compression techniques using a fixed block size cannot exploit the uniformity or redundancy of an image. To take advantage of this, we can vary the block size until it accommodates a certain minimum detail. Based on this, we present our novel variable block-size compression technique that estimates the detail of a block using edge detection. We consider representation of edges to be vital in the reconstruction of the image as they characterize sharp changes in the pixel values of the image. By classifying blocks into edges and nonedges using a criterion and merging only the nonedge blocks, the block-size is varied. The merged blocks are then transformed into a single block to form the source vector for quantization. This method represents the compressed image at a lower rate for the same amount of distortion. Satisfactory results are achieved when compared to fixed block-size quantization for bitrates between 0.1 and 0.6 bpp depending on the nature of the image.