Abstract
This thesis is about the evaluation of a prediction algorithm for a standard video codec proposed in [1]. This video codec follows the specifications of the ISO H.261 Recommendation. The video codec and the proposed algorithm were written in software, and the algorithm was implemented into the predictor of the video codec. Using Mean Field Annealing, the predictor iterates the reconstructed image to eliminate blocking artifacts, noise, and other image degradations. This is supposed to be achieved by using "a priori" information about the image to be iterated. The implemented video codec was run with and without the proposed algorithm on different PC platforms, and the results evaluated with various criteria. The results presented show that the proposed algorithm seems to be inappropriate for the intended application. Computing speed is slowed down considerably because of the iterative nature of the algorithm. Furthermore, the worse the quality of the considered image, the better results yields the algorithm. However , the resulting image quality is not good enough to justify an employment of this algorithm in real video telephony or conferencing systems. Other means of improving video image quality like reducing frame rates or reallocating bandwidth would be more beneficial.
Schwerdt, Karl (1996). Design and implementation of a video codex using Mean Field Annealing. Master's thesis, Texas A&M University. Available electronically from
https : / /hdl .handle .net /1969 .1 /ETD -TAMU -1996 -THESIS -S394.