论文部分内容阅读
为了使立体视频中的比特分配更加符合人眼视觉感知特性,提出了一种非对称质量的立体视频编码码率控制算法。首先,建立了左右帧的码率分配比例与量化参数差值之间的立体指数RRQ(Rate-ratio Quantization)模型。然后,将码率控制算法分为SGOP(Stereoscopic Group of Pictures)层、立体图像对层和帧层等3个码率控制层。在SGOP层计算每个SGOP的目标码率和关键帧的量化参数;在立体图像对层根据剩余比特数和缓冲区饱和度计算每个立体图像对的目标比特;在帧层则通过分析双目视觉掩蔽效应,用一种适合于立体视频的率失真优化方法合理分配左右帧的目标码率。实验结果表明,本文算法的码率控制偏差平均值为0.21%;立体视频客观质量比对称质量算法和Wang的算法分别提高了0.23dB和0.06dB,且质量波动较为稳定。因此,该算法基本满足网络带宽传输要求。由于充分利用了人眼双目视觉特性,可满足人们对立体视频的视觉需求。
In order to make the bit distribution in stereo video more accord with the visual perception of human eyes, an asymmetric quality stereo video coding rate control algorithm is proposed. First, the rate-ratio Quantization (RRQ) model is established between the rate allocation and the quantization parameter difference between the left and right frames. Then, the rate control algorithm is divided into three rate control layers such as a Stereoscopic Group of Pictures (SGOP) layer, a stereo image pair layer and a frame layer. The target bit rate of each SGOP and the quantization parameter of the key frame are calculated at the SGOP layer; the target bit of each stereo image pair is calculated at the stereoscopic image pair layer according to the remaining number of bits and the buffer saturation; Visual masking effect, using a rate-distortion optimization method suitable for stereoscopic video to rationally allocate the target bit rate of left and right frames. The experimental results show that the average deviation of the code rate control of this algorithm is 0.21%. The objective quality of stereo video is improved by 0.23dB and 0.06dB respectively compared with the symmetric and Wang algorithms, and the quality fluctuation is stable. Therefore, the algorithm basically meets the requirements of network bandwidth transmission. Due to the full use of the human eye binocular visual characteristics, to meet people’s visual needs of three-dimensional video.