Face image compression

In compression project, we are interested in investigation of special structured codebooks and coding techniques capable to achieve rate-distortion bounds with polynomial complexity.
The main focus of our research is on machine-learning based image compression applied to specific image datasets such as facial images, PUFs or astronomical images.


An Example of structured sparse codebook used for compression

Achievability of Shannon lower bound for the Gaussian source with the trained codebook (sequences are generated from the corresponding distribution) and binary distribution

Related publications

• S. Ferdowsi, S. Voloshynovskiy, and D. Kostadinov, "Sparse Multi-layer Image Approximation: Facial Image Compression,"ArXiv e-prints, 2015. [pdf|bib]
• S. Ferdowsi, S. Voloshynovskiy, D. Kostadinov, M. Korytkowski, and R. Scherer, "Secure Representation of Images Using Multi-layer Compression," , Rutkowski, L., Korytkowski, M., Scherer, R., Tadeusiewicz, R., Zadeh, L. A., and Zurada, J. M., Eds., Springer International Publishing, 2015, vol. 9119, pp. 696-705. [pdf|bib]
• J. E. Vila-Forcén, S. Voloshynovskiy, O. Koval, and T. Pun, "Facial Image Compression Based on Structured Codebooks in Overcomplete Domain," {EURASIP} Journal on Applied Signal Processing, Frames and overcomplete representations in signal processing, communications, and information theory special issue, vol. 2006, iss. Article {ID} 69042, pp. 1-11, 2006. [pdf|bib]
• S. Voloshynovskiy, O. Koval, and T. Pun, "Joint compression of independent modalities in smart meeting rooms applications," in Proc. SIMILAR NOE Workshop, Lausanne, Switzerland, 2004. [bib]
• J. E. Vila-Forcén, S. Voloshynovskiy, O. Koval, and T. Pun, "Low bit rate image compression using a non-stationary geometrical image priors," in Proc. IM2 Interactive Information Management Workshop, Martigny Crans-Montana, Switzerland, 2003. [bib]