The emerging video coding standard, H.264/MPEG-4 AVC, allows to use multiple reference frames for motion estimation to enhance temporal prediction. Exhaustive search of each frame requires a lot of computation which is propotional to the number of searched frames. However, the reduction of prediction residues is highly dependent on the characteristics of video sequences. In many cases, searching more reference frames contributes to nothing but only waste of computation. In this paper, we proposed a novel algorithm to accelerate the motion estimation by using reference frame skipping criteria. We adopted selected macroblock mode, intra/inter frame prediction residues, compactness of motion vectors, and scene changes in the criteria. By using these criteria, each macroblock can determine whether it is necessary to keep on searching more reference frams after the block matching process in the first reference frame. Simulation results show that the proposed algorithm can save up to 80% of computation without noticeable degradation of video quality.
[1]
Anil K. Jain.
Fundamentals of Digital Image Processing
,
2018,
Control of Color Imaging Systems.
[2]
J. Woods,et al.
Probability and Random Processes with Applications to Signal Processing
,
2001
.
[3]
Till Halbach.
CONCEPTS AND PERFORMANCE OF NEXT-GENERATION VIDEO COMPRESSION STANDARDIZATION
,
2002
.
[4]
Ming-Ting Sun,et al.
Modeling DCT coefficients for fast video encoding
,
1999,
IEEE Trans. Circuits Syst. Video Technol..
[5]
A. Desoky,et al.
Entropy measures for texture analysis based on Hadamard transform
,
1990,
IEEE Proceedings on Southeastcon.
[6]
Majid Rabbani,et al.
Biased reconstruction for JPEG decoding
,
1999,
IEEE Signal Processing Letters.
[7]
Jerry D. Gibson,et al.
Distributions of the Two-Dimensional DCT Coefficients for Images
,
1983,
IEEE Trans. Commun..