$\ell_{2}$ Restoration of $\ell_{\infty}$-Decoded Images Via Soft-Decision Estimation

摘要: The l∞-constrained image coding is a technique to achieve substantially lower bit rate than strictly (mathematically) lossless coding, while still imposing tight error bound at each pixel. However, this becomes inferior in the l2 distortion metric if decreases further. In paper, we propose new soft decoding approach reduce of l∞-decoded images and retain advantages both minmax least-square approximations. performed framework restoration that exploits bounds afforded by employs context modeler quantization errors. Experimental results demonstrate hard decoded can be restored gain more 2 dB peak signal-to-noise ratio PSNR, retaining on every single even outperform JPEG 2000 (a state-of-the-art encoder-optimized codec) for rates higher 1 bpp, critical region applications near-lossless compression. All gains are made without increasing encoder complexity as heavy computations efficiency delegated decoder.