Fast Walsh transform and multi-view video coding
-
摘要: 本文针对多视角视频编码提出了一种新的编码方法。在此方法中,结合四维Walsh操作算子,以达到压缩目的。利用4维n阶矩阵Walsh变换,对先前彩色视频流的编码加以扩展,将其应用到八个视角的视频编码中,包括视频序列分块,Walsh正变换及反变换,反分块。这种方法能够利用视频序列之间的相关性并且减少视频序列之间的冗余。本文以VC++6.0为工具,编程实现了基于快速Walsh变换的多视角视频编码,研究了不同压缩比条件下的压缩性能。通过对实验数据的分析,本文提出的方法既保证了视频质量又具有很好的快速压缩性能。实验结果表明:本文方法具有可行性及有效性,且易于在编码端快速实现,为多视角视频的进一步研究奠定了基础。Abstract: A fast coding method for multi-view video coding has been put forward in this paper. In this method, in order to compress the data, we combined four dimension Walsh operation. Combined with 4D n-order matrix Walsh transform, a series of coding schemes on colorful video stream proposed in previous studies was expanded and applied in eight-view video coding. The coding method includes video sub-blocking, Walsh transform and inverse transform, inverse sub-blocking, which can take advantage of the correlation of the video sequence and reduce the redundancy of the video sequence. We achieved the goal of multi-view video programming based on fast Walsh transform in VC++6.0 environment, and studied the compression performance in different compression conditions. Experiment data shows that the video quality can be guaranteed and the CR and PSNR is good. With good feasibility and effectiveness, this method is easy to achieve in the encoding side and lays a foundation for further study of multi-view video coding.
-
表 1 Walsh变换下不同分块方式的CR及PSNR
Table 1. CR and PSNR corresponding to different blockings in Walsh Transform
分块方式 CR PSNRy PSNRu PSNRv 8×8 8 38.27 39.43 38.76 16 35.87 36.13 36.25 16×16 8 36.29 35.88 36.13 16 34.14 33.14 33.87 表 2 DCT变换下不同分块方式的CR及PSNR
Table 2. CR and PSNR corresponding to different blockings in DCT Transform
分块方式 CR PSNRy PSNRu PSNRv 8×8 8 39.42 41.73 42.25 16 37.86 39.43 38.54 16×16 8 37.23 38.54 40.12 16 36.75 37.74 38.12 表 3 本文方法所用时间(ms)
Table 3. Time consumed in this method(ms)
分块方式 压缩比CR 分块 变换 反变换 反分块 总时间 8×8 8 5.32 6.33 6.30 5.14 23.09 16 4.89 5.85 5.79 4.71 21.24 16×16 8 4.67 5.48 5.39 4.55 20.09 16 4.24 5.12 4.93 4.09 18.38 -
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19]