TY - JOUR
T1 - Image and Video Compression Algorithms Based on Recovery Techniques Using Mean Field Annealing
AU - özcelik, Taner
AU - özcelik, Taner
AU - Katsaggelos, Aggelos K.
AU - Brailean, James C.
PY - 1995/2
Y1 - 1995/2
N2 - Image and video coding algorithms have found a number of applications ranging from video telephony on the Public Switched Telephone Networks (PSTN) to HDTV. However, as the bit rate is lowered, most of the existing techniques, as well as current standards, such as JPEG, H. 261, and MPEG-1 produce highly visible degradations in the reconstructed images primarily due to the information loss caused by the quantization process. In this paper, we propose an iterative technique to reduce the unwanted degradations, such as blocking and mosquito artifacts while keeping the necessary detail present in the original image. The proposed technique makes use of a priori information about the original image through a nonstationary Gauss-Markov model. Utilizing this model, a maximum a posteriori (MAP) estimate is obtained iteratively using mean field annealing. The fidelity to the data is preserved by projecting the image onto a constraint set defined by the quantizer at each iteration. The proposed solution represents an implementation of a paradigm we advocate, according to which the decoder is not simply undoing the operations performed by the encoder, but instead it solves an estimation problem based on the available bitstream and any prior knowledge about the source image. The performance of the proposed algorithm was tested on a JPEG, as well as on an H.261-type video codec. It is shown to be effective in removing the coding artifacts present in low bit rate compression.
AB - Image and video coding algorithms have found a number of applications ranging from video telephony on the Public Switched Telephone Networks (PSTN) to HDTV. However, as the bit rate is lowered, most of the existing techniques, as well as current standards, such as JPEG, H. 261, and MPEG-1 produce highly visible degradations in the reconstructed images primarily due to the information loss caused by the quantization process. In this paper, we propose an iterative technique to reduce the unwanted degradations, such as blocking and mosquito artifacts while keeping the necessary detail present in the original image. The proposed technique makes use of a priori information about the original image through a nonstationary Gauss-Markov model. Utilizing this model, a maximum a posteriori (MAP) estimate is obtained iteratively using mean field annealing. The fidelity to the data is preserved by projecting the image onto a constraint set defined by the quantizer at each iteration. The proposed solution represents an implementation of a paradigm we advocate, according to which the decoder is not simply undoing the operations performed by the encoder, but instead it solves an estimation problem based on the available bitstream and any prior knowledge about the source image. The performance of the proposed algorithm was tested on a JPEG, as well as on an H.261-type video codec. It is shown to be effective in removing the coding artifacts present in low bit rate compression.
KW - Image and video compression
KW - image and video coding
KW - mean field annealing
UR - http://www.scopus.com/inward/record.url?scp=0029247211&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=0029247211&partnerID=8YFLogxK
U2 - 10.1109/5.364460
DO - 10.1109/5.364460
M3 - Article
AN - SCOPUS:0029247211
SN - 0018-9219
VL - 83
SP - 304
EP - 316
JO - Proceedings of the IEEE
JF - Proceedings of the IEEE
IS - 2
ER -