一种基于视频的超分辨率图像生成方法研究

A Super Resolution Image Construction Method Based on Video Clips

随着手持视频设备的普及,如何为这类低分辨率应用环境提供一种生成高质量图像的方法是一个值得研究的问题.讨论了如何从低分辨率、低质量的视频中创建清晰的、高分辨率图像的方法.针对当前多图像超分辨率技术没有考虑运动估计和重要帧的权值等问题,提出了一个基于时间显著性、局部图像统计特征来计算图像像素权值的方法,以实现更好的图像融合.提出的基于重要性的计算框架,降低了低质量帧的影响,显著提高了有效成分的作用.

With the popularity of mobile video devices,the problem of how to produce high resolution and high quality images for those mobile devices which only support low resolution and quality image deserves researchers＇ attention. This paper studies how to make high resolution and high quality images from the low resolution and quality ones. Taking motion estimation and weight of significant frames,that are often ignored in the literature,into consideration,we develop an image pixed weight calculation scheme based on time significance and local image statistics features to better fuse the image. The proposed framework based on significance could effectively increase the influence of the important images,while decrease the effect of the low quality images or blocks.