The Study Of Fast Algorithms For Scalable Video Coding

Scalable video coding (SVC) is currently a hot research topic of video coding. Due to the limited network bandwidth available and the enormous amount of video data, videos need to be compressed for transmission by the techniques called video coding. Further, with the development of video coding and network technologies, the multimedia applications are running on an environment consisted of heterogeneous networks and diverse terminals. This requires the bit stream generated by the coding method to be adaptive to such situation, and gives rise to the SVC technologies, which are more efficient than the previous standards at the cost of increasing the coding complexity incurred by the introduction of Inter-Layer Prediction. Therefore, how to reduce the coding complexity and coding time, and increase the coding speed without significantly sacrificing the coding efficiency is a very important research direction in video coding.This paper explores the current research on the SVC technologies and their deficiencies in a comprehensive and systematical way, and proposes a serial of related improvement and optimization algorithms.The core content and novelties of this paper are as the follows:(1)This paper analyzes the principles and coding schemes of the SVC technologies, and paves the way for the fast coding algorithms introduced later.(2) This paper proposes a temporal scalable intra-frame fast coding algorithm, which is suitable for the Intra8 mode and proposed based on the observation that the Intra8 mode newly introduced into the standard intra-frame coding algorithms improves the coding efficiency significantly at the cost of increasing the coding complexity.(3) This paper proposes a temporal scalable inter-frame fast coding algorithm. The scalable video coder uses the hierarchical B-Pictures structure to achieve the temporal scalability with the limitation of not suitable for the hierarchical B-Pictures structure. Therefore, this paper proposes an inter-frame fast coding algorithm suitable for the hierarchical B-Pictures structure based on the in-depth analysis of the principles of the inter-frame mode selection and the reference model, (4)This paper proposes a fast enhancement layer intra-prediction for the SVC algorithm. This paper speeds up the encoding speed by analyzing the relationship between the base and enhancement layers, and also that between the left-adjacent and the top adjacent block, and proposing the block mode selection and the prediction of the residual coefficients. The speed is further increased by the design of the direction modes of selection in Intra16 and Intra4.(5)This paper proposes the inter-frame fast coding algorithm for the spatial SVC enhancement layer. We speed up the encoding speed by the method, in which the weight of the macro block is assigned by its related encoding method, and then encoded in an order according its weight value, combined with the early termination using the residuals prediction. The sub-block mode of the 8×8 sub-block is calculated through the hypothesis testing. For the modes weighted by zero, the encoding speed is speeded up further by the value of the rate-distortion prediction.(6)This paper proposes a fast encoding algorithm. The encoding speed is increased by first carrying out the hypothesis testing using the residual coefficients obtained from the Inter16×16 or Inter8×8. Then the results of the hypothesis testing and the rate-distortion value are applied to predict the possible coding modes by eliminating the ones with lower possibility. By avoiding some or even all of the operations on the coefficients of DCT, Q, IQ, IDCT in the 4×4 sub-blocks using the criteria for all-zero blocks, 1×1 sub-blocks, 2×2 sub-blocks and 3×3 sub-blocks calculated based on the distribution of the DCT coefficients, further improvement on the encoding speed is also achieved.