Abstract:
In this thesis a novel, low memory P-Skip mode decision algorithm for encoding video sequences is proposed. The existing algorithms for P-Skip
estimation require reading the previous frames from the memory. The proposed algorithm saves signi cant memory by eliminating the need for storing the reference frame by maintaining an optimal 'Similarity Metric (SM)'
for each macroblock. P-Skip decision is taken on the basis of the correlation between the similarity metrics of co-located macroblocks in consecutive frames. A Weighted Matrix (WM) based SM is chosen in this thesis.
Also, the signi cance of the objective Video Quality Assessment (VQA)
metrics, like Peak-Signal-to-Noise Ratio (PSNR) and Structural Similarity
Index (SSIM) in nding the artifacts at the macroblock level, is gauged in
this thesis. Experimental results show that a signi cant memory reduction
of 97.46% is achieved with a 0.193 dB increase in the PSNR (quality) at
the cost of 0.3368% increase in bitrate. Application areas include Wireless
Gigabit (WiGig) use cases, low cost video calls and sharing of synthetic
content over the Internet.