基于分区曝光融合的不均匀亮度视频增强

发布时间：2019-01-06 16:35

【摘要】：视频增强技术可以使低曝光视频中的细节显示出来,有效地改进视频质量,为视频的后期处理提供保证.针对直接将不均匀亮度图像增强方法运用到视频处理中容易出现时空不一致现象的问题,提出一种基于分区曝光融合的不均匀亮度视频增强方法.为了实现视频帧的均匀曝光,首先对输入视频帧采用分区曝光法计算不同的曝光区域,对不同的曝光区域进行多尺度曝光校正;然后对每一个区域融合其不同曝光程度的区域;最终形成完整的曝光良好的视频帧.对于视频帧之间时空一致性的保持,结合前一帧曝光分区的方法能更加准确地计算相邻帧之间的对应关系,将前一帧的相应的像素传播到后面的帧中.实验结果表明,相比于目前的视频增强方法,该方法可以使不均匀亮度视频均匀曝光并且更好地保持视频场景的自然性,同时保持视频的时空一致性.
[Abstract]:Video enhancement technology can make the details of low exposure video display, improve the quality of video effectively, and guarantee the post-processing of video. Aiming at the problem that the non-uniform luminance image enhancement method is easy to occur spatio-temporal inconsistency in video processing, a new method of non-uniform brightness video enhancement based on partition exposure fusion is proposed. In order to achieve uniform exposure of video frames, different exposure regions are calculated by partition exposure method for input video frames, and multi-scale exposure correction is carried out for different exposure regions, and then each region is fused with different exposure areas. Finally, a complete video frame with good exposure is formed. For the maintenance of temporal and spatial consistency between video frames, the corresponding relationship between adjacent frames can be calculated more accurately by combining the previous frame exposure partition method, and the corresponding pixels of the previous frame can be propagated to the following frame. The experimental results show that compared with the current video enhancement method, the proposed method can make the inhomogeneous brightness video exposure uniformly and preserve the naturalness of the video scene, while maintaining the temporal and spatial consistency of the video.
【作者单位】：天津大学计算机科学与技术学院;
【基金】：国家自然科学基金(61170118,61672375)
【分类号】：TP391.41

【参考文献】