Regularized dequantizers for DCT-based transform coding of images
Gunho Lee, Sinae Kim, Samuel Moon-Ho Song · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2000
We have recently proposed a new dequantization scheme for DCT- based transform coding based on regularization principles. The new approach sharply reduced blocking artifacts in decoded images and the performance of the new dequantization scheme has been evaluated against the standard JPEG, MPEG and H.263+ in terms of the peak-signal-to-noise ratio (PSNR) with our own definition of the blockiness measure (BM). Basically, the proposed dequantizer maps the received data to within the range +/- (quantizer spacing/2), so that the final decompressed image is 'smooth' in the sense of minimizing the cost functional including the stabilizing term weighted by a regularization parameter. In this paper, we focus on several important aspect of this regularized dequantizer, namely the selection of the regularization parameter and the convergence of the dequantization algorithm.