基于感兴趣区域的H.265样点自适应补偿算法优化

针对目前H.265环路滤波器样点自适应补偿(sample adaptive offset,SAO)算法复杂度高、计算量大的问题,提出了一种基于人眼感兴趣区域的样点自适应补偿技术.人眼感兴趣区域在视频的空间域对应纹理复杂的区域,在时间域对应运动剧烈的对象.利用图像的时域和空域的二维信息,使用编码单元深度和运动矢量分别检测纹理复杂区域和运动剧烈的对象,结合深度和运动矢量将视频帧划分为人眼不感兴趣、较感兴趣、很感兴趣3种类型,并分别采用不补偿、弱补偿和强补偿的SAO算法,从而在总体上减少自适应补偿的计算量,提高编码效率.实验表明,所提算法可使样点自适应补偿编码时间平均减少61.47％,而峰值信噪比(peak signal-to-noise ratio,PSNR)仅有0.033 dB的降低.

Optimization of H.265 sample adaptive offset algorithm based on region of interest

A novel method is proposed in order to reduce the high complexity of the sampling adaptive offset (SAO) technology in the loop filter of H.265. The method introduces an adaptive compensation algorithm for different video frames by analyzing the regions of interest of human eyes, which are generally represented by complex texture areas in the spatial domain, and objects with intense motions in the temporal domain. The algorithm uses both the time domain and the spatial domain information of the image to calculate the depth and motion vectors, detects the complex texture regions and the motion regions accordingly. Then, the algorithm uses both the depth and motion vectors to determine the interest intensity of human eyes to the images, and divides the video frames into three types:"non-interested", "weakly interested", and "highly interested", and adopts the method of no compensation, weak compensation and high compensation to reduce the calculation of adaptive compensation and improve the coding efficiency. The experimental results show that the proposed algorithm can reduce about 61.47% of the average time of SAO coding, while the PSNR of video quality only decreases by 0.033 dB.