Coding artifact reduction of JPEG images using SSIM-optimal filtering with quantization constraint
Yosuke Shigedomi, Kaiyu Suzuki, Iciro Matsuda · 2024
Since the JPEG standard employs discrete cosine transform (DCT) based lossy coding algorithm, quality degradation due to coding artifacts is unavoidable especially at low coding rates. In order to alleviate this problem, we previously proposed a post filtering method that can maximize the well-known objective quality index called SSIM. In this paper, we combine the method with the projection onto convex sets (POCS) algorithm to exploit quantization constraint (QC) prior on DCT coefficients extracted from the JPEG image. Experimental results indicate that SSIM scores of the JPEG images are improved by applying the POCS algorithm.