| 摘要: |
| 为了更好地将人眼视觉感知特性应用于视频压缩系统，从而去除视频的视觉冗余成分，在利用像素域的恰可察失真(Just Noticeable Distortion,JND)阈值的基础上，结合JND阈值的空间相关性与图像块类型，提出了一种改进的残差滤波算法。该算法首先在像素域计算出每个像素的JND阈值，然后在变换单元(Transform Unit,TU)中挖掘每个像素JND阈值与其周围邻近像素JND阈值之间的空间相关性，再利用索贝儿(Sobel)边缘检测算子将TU分成不同类型的图像块，并且计算出对应的复杂度因子，最后结合上述像素JND阈值的空间相关性和TU复杂度因子对TU残差进行滤波。提出的算法模型可以嵌入到高效率视频编码(High Efficiency Video Coding,HEVC)框架。实验结果表明，在全I帧配置下，提出的算法与标准算法HM16.0相比，在人眼主观感知质量基本一致的情况下，平均可节省16.1%的码率。 |
| 关键词: 视频压缩系统；视觉感知特性 恰可察觉失真（JND） 高效率视频编码(HEVC) 残差滤波 |
| DOI： |
|
| 基金项目:国家自然科学基金资助项目(61871279，61471248)；成都市科技惠民技术研发项目（2015-HM01-00293-SF）；成都市产业集群协同创新项目（2016-XT00-00015-GX） |
|
| A residual filtering algorithm combining pixel domain spatial correlation of JND threshold and image block type |
| ZHANG Feiran,HE Xiaohai,QING Linbo,TIAN Gang,XIONG Shuhua,WU Xiaoqiang |
| (College of Electronics and Information Engineering,Sichuan University,Chengdu 610065,China;Administrative Affairs Center，Xinjiang Oilfield Company，Karamay 834000，China) |
| Abstract: |
| In order to better apply the human visual perception property to video compression system to remove the visual redundancy of video,a residual filtering algorithm is proposed which combines the spatial correlation of just noticeable distortion(JND) threshold in pixel domain and image block type.Firstly,the JND threshold is calculated for each pixel in pixel domain，and then the spatial correlation between each pixel's JND threshold and its neighboring pixel's JND threshold is discovered in transform unit(TU).Then the Sobel edge detector is applied to classify the TU into different image block types and the complexity factor is calculated according to TU's type.Finally,the TU's residual is filtered by combining the spatial correlation of pixel's JND threshold and the TU's complexity factor.The proposed algorithm model can be embedded into the high efficiency video coding(HEVC) framework.The experimental results show that under all I-frame configuration the proposed algorithm can obtain average 16.1% bit rate reduction for HEVC while keeping nearly the same subjective perceptual quality. |
| Key words: video compression system visual perception characteristics just perceptible distortion(JND) high efficiency video coding(HEVC) residual filtering |
