论文部分内容阅读
为了达到减少比特数同时保持画面质量的目的,提出了一种基于最小可视失真(JND)和自回归(AR)模型的感知视频编码方法.首先,设计了基于JND的纹理分割算法,建立了空时JND模型,以MB为基本单元,通过计算其JND能量并与阈值做比较,用以分割出视频序列中的纹理区域.然后,开发了AR模型来合成纹理区,在使用最小二乘法计算出AR模型的参数后,用相邻的前后参考帧对应像素的线性插值来生成重构像素.最后,为了检验所提方法的效果,将其与H.264/AVC视频编码系统做比较,用不同的视频序列实验来验证所提方法的有效性.实验结果显示,对于具有不同纹理特点的实验序列,所提方法可以在保持感知质量的同时将比特率减少15%～58%.
In order to reduce the number of bits and maintain the picture quality, a perceptual video coding method based on minimum visual distortions (JND) and autoregressive (AR) models is proposed.Firstly, a texture partitioning algorithm based on JND is designed and a The space-time JND model takes MB as the basic unit, and computes its JND energy and compares it with the threshold to segment the texture region in the video sequence.And then, AR model is developed to compose the texture region, which is calculated by the least square method After the parameters of AR model are obtained, the reconstructed pixels are generated by the linear interpolation of the corresponding pixels of the adjacent reference frames.Finally, in order to test the effectiveness of the proposed method, this method is compared with H.264 / AVC video coding system Different video sequence experiments to verify the effectiveness of the proposed method.The experimental results show that for experiments with different texture features, the proposed method can reduce the bitrate by 15% -58% while maintaining the perceived quality.