The Research Of Video Transcoding For Spatial Resolution In H.264

Video transcoding is the operation of converting a compressed video stream to the other stream of different form for adapting to the network and usage environment. The forms include the coding syntax, code rate, frame rate and spatial resolution. With the extensive use of the network appliances including handheld computers, personal digital assistants (PDAs), MP4, and smart cellular phones, which varies significantly in resources such as computing power and display capability. To provide video contents to these appliances, video transcoding is necessary. Among them, spatial transcoding is one of the most important kinds of transcoding. H.264 is the latest video coding standard, so this thesis focus on the H.264 related spatial transcoding.This paper first introduces the conception and classification of transcoding, which includes the conversion of bit-rate, spatial resolution, frame-rate and the code syntax. The features, key techniques and the search hot-topics of every category are briefly introduced. The main contributions of this thesis are detailed as follows: First, the architectures for spatial transcoding are surveyed, and the existing problems and key techniques for implementation are summarized. Second, a motion re-estimation algorithm based on hybrid rectangle and hexagon templates is proposed. It utilizes the motion vectors, which are easily obtained from the input video stream, to derive the directional derivative and gratitude, and then predicts the initial motion vectors. Motion re-estimation algorithm selects different search pattern of rectangle and hexagon template to obtain the proper the motion vectors. Third, a local direction based fast mode selection algorithm is proposed. For INTRA model selection, it utilizes the histogram of the local prediction direction and selects that of high probability. For INTER mode selection, it is based on the gradient to reduce the number of candidate modes. Experimental results demonstrate that it can reduce the complexity of mode decision in downscaling video transcoding while keeping the quality of reconstructed video and coding efficiency.At last, future work worthy of further investigation in this field is prospected.