多视点视频中基于乘加误差模型的亮度和色差校正

Luminance and chrominance correction for multi-view video using multiplicative and additive errors

由于相机性能的差异,多视点视频序列之间总是存在亮度和颜色的差异,这种差异会影响多视点视频压缩和显示等后续处理过程的性能。本文由此提出了一种适用于多视点视频的亮度和色差(亮色度)校正算法。根据相机成像原理,不同相机的亮色度差异可以归类为乘性误差和加性误差。灰度级偏移,相机电子元件的放大和量化因子差异等都可以由乘加误差模型表示。为降低计算复杂度,本文采用直方图形状匹配的方法计算乘加误差模型的参数。该算法可作为多视点视频系统中的预处理过程,实验结果表明,该算法能够有效地提高后续压缩过程的性能。

Significant luminance and chrominance discrepancies between the different camera views can often be observed in multiview video sequences. These variations may well impair the performance of a multi-view coder or a renderer. A luminance and chrominance correction algorithm is proposed in this paper. According to the theory of image formation, luminance and chrominance variation are categorized into multiplicative and additive errors, which can express various error sources such as gray-level shift, variations in amplification and quantization in camera electronics, and related factors. Histogram matching is used to compute multiplicative and additive variation factor for its low computational complexity. The proposed algorithm can be used as preprocessing tool in the multiview video system. Experiment results show that it can efficiently improve the performance of multiview video encoder.