基于SSE和SSIM的H.264帧内预测模式选择改进算法

摘要

在H.264的帧内预测模式选择过程中,率失真优化(RDO)的失真测度用当前编码块和预测块之间的平方误差和(SSE)或绝对误差和(SAD)来度量,而这两种失真测度被证明不能很好地符合人眼视觉(HVS).该文参考软件JM16.2,提出了联合SSE和结构相似度(SSIM)作为失真测度用于RDO的方法(CSSRDO).算法首先找到SSIM和码率的近似关系,然后综合以SSE作为失真测度的RDO函数,并结合人眼视觉特性,建立了联合SSE和SSIM作为失真测度的RDO模型.实验表明,将CSSRDO用于H.264帧内预测模式选择获得了比JM16.2更高的编码效率和更好的重建图像质量.%The distortion metrics Sum of Square Error (SSE) and Sum of Absolute Difference (SAD), adopted in RDO on H.264 reference software, have been proved not correlating well with Human Visual System (HVS).Referring to JM16.2, this paper proposes to Combine SSE and Structural SIMilarity (SSIM) as the distortion metric in RDO (CSSRDO) instead of SSE. Firstly, the algorithm finds the appropriate relationship between SSIM and rate. After utilizing the RDO function where distortion is measured as SSE and considering the HVS, the RDO based on combined SSE and SSIM aa the distortion metric is proposed. Further more, CSSRDO is introduced into JM16.2 for intra-mode selection. Results show that the proposed algorithm can achieve better coding efficiency and image quality.