A study on the video compression using the image matching based on the critical-point filters

我们提出了一种通过首先找到关键帧之间的密集对应关系，然后通过压缩残差来压缩视频序列等图像序列的方法。这项拟议工作的核心是使用一种名为CPF的自动图像匹配算法，该算法在任意图像之间建立逐个像素的映射。不需要关于图像中的对象的任何先验知识。参数完全通过动态计算自动设置。CPF在每个分辨率下计算两幅图像之间的最优映射，其中多分辨率层次由保留较强临界点的CPFS生成。因此，CPF适用于对图像内容没有任何假设的任意图像序列。

We have proposed a method to compress a sequence of images such as a video sequence by first finding the dense correspondence between keyframes and then by compressing the residuals. Central to the proposed work is the use of an automatic image matching algorithm called CPF that establishes a pixel-by-pixel mapping between arbitrary images. There is no need for any prior knowledge of the objects in the images. Parameters are set completely automatically by dynamical computation. CPF computes the optimal mapping the between two images at each resolution where the multiresolution hierarchy is generated by the CPFs that preserve the stronger critical points. Thus, CPF is suitable for arbitrary image sequences where there can be no assumption on the image contents.