Fast prediction algorithm for multiview video coding

The H.264/multiview video coding (MVC) standard has been developed to enable efficient coding for three-dimensional and multiple viewpoint video sequences. The interview statistical dependencies are utilized and an interview prediction is employed to provide more efficient coding; however, this increases the overall encoding complexity. Motion homogeneity is exploited here to selectively enable interview prediction, and to reduce complexity in the motion estimation (ME) and the mode selection processes. This has been accomplished by defining situations that relate macro-blocks' motion characteristics to the mode selection and the interview prediction processes. When comparing the proposed algorithm to the H.264/MVC reference software and other recent work, the experimental results demonstrate a significant reduction in ME time while maintaining similar rate-distortion performance. 1 Introduction Recent advances in digital technologies have paved the way for the development of numerous real-time applications deemed too complex in the past. A vast array of those applications requires transmission and storage of digital videos. Examples include, but are not limited to, digital TV, video streaming, multimedia communications, remote monitoring, videophones and video conferencing. Advances in digital video can be classified as one of the most influential modern technologies, due in part to the fast, widespread use of digital video applications in everyday life. Consequently, over the last three decades, high-quality digital video has been the goal of companies, researchers and standardization bodies. 1 Multiview, in particular, is a highly relevant topic in the current research by institutions attempting to advance real-time multiview applications. As a result, and due to the wide expansion of three-dimensional (3-D) and free viewpoint video applications, the H.264 multiview video coding (MVC) standard has been developed as an extension to the H.264/ advanced video coding (AVC) standard to enable efficient coding of scenes captured from multiple cameras. 2 Since all cameras capture the same scene from different viewpoints, interview statistical dependencies can be expected. Therefore , in addition to the H.264/AVC very refined motion estimation (ME) and motion compensation (MC) processes, H.264/MVC exploits interview prediction for more efficient coding. However, this further increases the overall encoding complexity and makes the standard complexity unbearable for real-time encoding. Therefore, there is a tremendous need to reduce encoding complexity and to design a flexible, rate-distortion-optimized, yet computationally efficient, encoder for various applications. Recently proposed algorithms reduce encoder complexity by locating corresponding objects in neighboring views by means of a global disparity vector and exploiting the mode