An Analytic Framework for Frame-Level Dependent Bit Allocation in Hybrid Video Coding

摘要

In this paper, we address the frame-level dependent bit allocation (DBA) problem in hybrid video coding. In most existing methods, the DBA solution is achieved at the expense of high, sometimes even unbearable, computational complexity because of the multipass coding involved. Motivated by this, we propose a model-based approach as an attempt to solve this problem analytically. Leveraging the predictive nature in hybrid video coding, we develop a novel interframe dependency model (IFDM), which enables a quantitative measure of the coding dependency between the current frame and its reference frame. Based on the IFDM, the buffer-constrained frame-level DBA problem is carefully formulated. Finally, the model-based DBA method called IFDM-DBA is derived, in which successive convex approximation techniques are employed to convert the original optimization problem into a series of convex optimization problem s of which the optimal solutions can be obtained efficiently. Experimental results suggest that the proposed IFDM-DBA method can achieve up to a 23% bitrate reduction over the JM reference software of H.264.