Research Of Fast Algorithm For Intra Prediction And The Integer Transform In HEVC

HEVC (High Efficiency Video Coding) is the new generation of highly efficient video coding standard, its purpose is to enhance video compression efficiency, and adapt to different network environment. With aim at improving the encoding precision and performance, the kinds of HEVC intra prediction modes is increased from 9 to 35, and the kinds of transform block size is increased from 1 (4×4) to 4 (4×4,8x8,16x16, and 32x32) compared to H.264/AVC. However, the increase of the mode sorts and the size also brings higher computational complexity. For these two problems, this paper presents two kinds of fast algorithms in order to improve the encoding efficiency. The main work is shown as following:1. In order to reduce the number of candidate patterns in the Rough Mode Decision (RMD) and speed up the frame of mode selection, this paper proposes a fast mode decision algorithm with a combination of image texture direction and spatial correlation in HEVC. In the algorithm, by using the Sobel operator to obtain the texture direction of the current block Prediction Union (PU), the corresponding angle prediction model is selected to compose the RMD candidate list; and by using the spatial correlation of images, the optimal mode of the neighbor PU is added to the candidate list. This algorithm will reduce the number of candidate patterns from 35 to less than 10 in the RMD. The experimental results show that this method significantly shorten the encoding time of 34.9% with the bit-rate increasing 0.41% and the PSNR falling 0.0309dB,.2. In order to improve the computing speed of DCT transformation, a fast DCT algorithmof divided subbands based on HEVC is proposed. According to the properties of the characteristics of "Z scan"and DCT coefficients’ energy distribution, DCT coefficients of the prediction residues are predicted to zero before implementing DCT and quantization (ZQDCT). One adaptive scheme is also presented with multiple thresholds to divide the subbands. By this scheme, only the DCT coefficients without ZQDCT will be computed. The experimental results show that this method significantly shorten the DCT time of 12.95% with the bit-rate increasing 0.24% and the PSNR falling 0.0203dB.In this dissertation, we draw insight into the intra prediction and DCT transform of the state-of-the-art HEVC video coding standard. Under the condition that the encoding quality is invariant, this method significantly reduces the complexity of encoding algorithm and improves the HEVC coding efficiency.