基于H.264的复杂度可分级的DCT算法研究

Complexity scalable DCT algorithm based on H.264

H.264标准在数字视频通信和存储领域有很广泛的应用市场,当其应用于手机等资源受限的平台上时,要求其编码器计算复杂度灵活可控,具有很好的可分级性。本文针对H.264整数变换、量化和Zig-zag扫描的特点,基于DCT高频系数截断思想提出了复杂度可分级的DCT算法,该算法通过调整复杂度控制参数来实现DCT模块复杂度的可分级。实验结果表明,该算法随着DCT分级模式的增加,计算复杂度、编码比特率和PSNR值都会下降,DCT分级模式为5时DCT模块的复杂度仅为分级模式为0时复杂度的18.75%,但图像整体主观质量变化不大。

H. 264 standard has a broad application in digital video communication and storage areas. When it is applied on resource constrained platforms such as handset,which requires that the computational complexity of the encoder can be controlled flexibly and has good complexity scalability. Therefore, taking account of the characteristics of integer transform, quant and zig-zag scan in H. 264, a complexity scalable DCT arithmetic has been produced based on the DCT pruning thought. The algorithm realizes the complextiy scalability of DCT module by adjusting complexity control parameters. Experimental results show that the proposed methods can bring the decrease of computational complexity, encoding bit-rate and PSNR with the increase of DCT classification modes, but the subjective qualtiy of images has little change. When DCT classification mode is 5, the complexity of DCT module is 18. 75% of the case when classificaiton mode is 0.