JPEG estimated spectrum adaptive postfilter using image-adaptive Q-tables and Canny edge detectors

The Estimated Spectrum Adaptive Postfilter (ESAP) combines several techniques to improve both the objective and subjective quality of low bit-rate JPEG gray-scale images. ESAP uses image-adaptive DCT Quantization Matrices for image encoding. After decoding, ESAP estimates 2-D pixel-adaptive bandwidths directly from the dequantized DCT coefficients. It then combines this information with directional Canny edge detectors to control a 2-D spatially-adaptive non-linear postfilter h[n₁, n₂] that significantly reduces the DCT blocking artifacts. ESAP exploits the Human Visual System (HVS) spatial frequency masking characteristics. Based on the HVS tolerance to quantisation errors in the high frequency regions, ESAP performs directional filtering parallel to the edges and no filtering across the edges, subject to filter design constraints. It adaptively filters low frequency regions to minimize blocking and restores the natural smoothness without noticeably reducing the fidelity of the true edges of the image. The non-edges are postfiltered with separable non-directional adaptive lowpass filters while the edges are postfiltered with non-separable directional lowpass filters. ESAP generally improves the PSNR 1 to 2.75 dB over baseline JPEG, and yields subjective improvement as well. ESAP relies on a DFT analysis of the DCT and is compliant with the Independent JPEG Group (IJG) Version 5b Release