Fast Bayesian JPEG Decompression and Denoising With Tight Frame Priors

0471741 - ÚTIA 2018 RIV US eng J - Journal Article
Šorel, Michal - Bartoš, Michal
Fast Bayesian JPEG Decompression and Denoising With Tight Frame Priors.
IEEE Transactions on Image Processing. Roč. 26, č. 1 (2017), s. 490-501. ISSN 1057-7149. E-ISSN 1941-0042
R&D Projects: GA ČR(CZ) GA16-13830S
Institutional support: RVO:67985556
Keywords : image processing * image restoration * JPEG
OECD category: Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
Impact factor: 5.072, year: 2017
http://library.utia.cas.cz/separaty/2017/ZOI/sorel-0471741.pdf

JPEG decompression can be understood as an image reconstruction problem similar to denoising or deconvolution. Such problems can be solved within the Bayesian maximum a posteriori probability framework by iterative optimization algorithms. Prior knowledge about an image is usually described
by the l1 norm of its sparse domain representation. For many problems, if the sparse domain forms a tight frame, optimization by the alternating direction method of multipliers can be very
efficient. However, for JPEG, such solution is not straightforward, e.g., due to quantization and subsampling of chrominance channels. Derivation of such solution is the main contribution of this paper. In addition, we show that a minor modification of the proposed algorithm solves simultaneously the problem of image denoising. In the experimental section, we analyze the behavior of the proposed decompression algorithm in a small number of iterations with an interesting conclusion that this mode outperforms full convergence. Example images demonstrate
the visual quality of decompression and quantitative experiments compare the algorithm with other state-of-the-art methods.
Permanent Link: http://hdl.handle.net/11104/0270739