Analysis and Improvement of JPEG Compression Performance using Custom Quantization and Block Boundary Classifications

JPEG (Joint Photographic Experts Group) compression is the global standard for digital image compression introduced in 1992, and is still wide spread use. However, at low bitrates the JPEG process can introduce unwanted visual artifacts such as the blocking effects or edge ringing. This paper describes a method for modification and customizing of the JPEG compression. A nonlinear relationship between the quantization matrix, reflecting the compression ratio and the peak signal-to-noise ratio (PSNR), as an objective quality measure, was experimentally determined. The estimation of the quantization matrix and approximation of its mapping to a PSNR is accomplished relying on transformation of eleven test images using all quantization matrices. The linear approximation to this relation in the region of interest was proposed enabling fine tuning of the reconstructed image quality by either selection of the desired PSNR value or a decompressed image quality. In the decompression phase, post-processing is applied to reduce the blockish artifacts introduced by the compression process. The image block boundaries are first classified for an automatic identification of high blockiness, to constrain the application of a pre-processing algorithm and further loss of an image detail. Upon the reconstruction, the quality of the reconstructed image is measured using the PSNR and structural similarity index (SSIM). The effects of compression on the spectral properties are analyzed by comparison of the original and decompressed image spectra.