基于JND与梯度显著度的HDR视频编码方法  

HDR video coding method based on JND and gradient saliency

作　　者：赵旭辉 郁梅[1,2] 徐海勇[1] 宋洋[1] ZHAO Xuhui;YU Mei;XU Haiyong;SONG Yang(Faculy of Information Science and Engineering,Ningbo University,Ningbo Zhejiang 315211,China;National Key Lab of Software New Technology,Nanjing University,Nanjing 210093,China)

机构地区：[1]宁波大学信息科学与工程学院,浙江宁波315211 [2]南京大学计算机软件新技术国家重点实验室,南京210093

出　　处：《激光杂志》2020年第7期72-79,共8页Laser Journal

基　　金：国家自然科学基金(No.61671258)。

摘　　要：高动态范围(High Dynamic Range,HDR)成像能够呈现更加真实的画面,由于其庞大的数据量,如何结合视觉感知实现其高效编码十分重要。提出一种基于恰可察觉失真(Just Noticeable Difference,JND)与梯度显著度(Gradient Saliency,GS)的HDR视频编码方法。首先,依据最大熵定理分割HDR视频帧得到高亮、低暗与中间亮度区域,根据分割结果计算并修正针对HDR视觉内容的JND阈值。然后,出于不同复杂程度的纹理信息对于人眼显著性不同的考虑,使用GS作为人眼获取视觉信息的影响因素。最后,将GS协同JND通过感知模型,进而指导HDR视频编码的率失真优化过程,实现在HDR视频编码中更加符合视觉感知的码率分配并降低码率。实验结果表明,所提方法能够有效提高HDR视频编码性能。High dynamic range(HDR)imaging can present a more real picture.Because of its huge amount of data,how to combine with the visual perception to realize efficient coding is very important.This paper proposes an HDR video coding method based on just noticeable difference(JND)and gradient salience(GS).Firstly,according to the maximum entropy theorem,HDR video frames are segmented to get high,low and middle brightness regions,and the JND thresholds for HDR visual content are calculated and modified according to segmentation results.Then,considering the different complexity of texture information for different significance of human eyes,GS is used as the influencing factor for human eyes to obtain visual information.Finally,GS and JND are used to guide the rate distortion optimization process of HDR video coding through the perception model,so as to realize the bit rate allocation and reduce the bit rate more in line with visual perception in HDR video coding.Experimental results show that the proposed method can effectively improve the performance of HDR video coding.

关 键 词：高动态范围视频 视频编码 率失真优化 梯度显著度 恰可察觉失真 

分 类 号：TN919[电子电信—通信与信息系统]