单尺度SSR

非局部全变分小波域图像修复代码，实现算法WAVELET INPAINTING BY NONLOCAL TOTAL VARIATION，Xiaoqun Zhang，Tony F. Chan，Inverse Problems and Imaging，2010,4（1）(Non-local full variational wavelet domain image repair code algorithm WAVELET INPAINTING BY NONLOCAL TOTAL VARIATION Xiaoqun Zhang, Tony F. Chan, Inverse from and Imaging, 2010,4 (1))

说明：  水下图像处理方法之一的图像融合算法matlab实现(Matlab implementation of image fusion algorithm, one of the classical image processing methods)

多尺度快速角点检测算法（FAST).Adaptive and generic corner detection based on the accelerated segment test. 出处 Computer Vision–ECCV 2010(Multi-scale fast corner detection algorithm (FAST) Adaptive and generic corner detection based on the accelerated segment test. Computer Vision-ECCV 2010)

rk3399硬件编解码demo及文档

该资料中的程序实现的是图像压缩，供初学者学习。(The data of the program implementation is the image compression, for beginners to learn.)

该代码用于实现视觉目标跟踪研究中的协方差跟踪，能够把多种时空特征融合于统一的模型中，在实现视觉目标跟踪时具有较好的鲁棒性，而且其维数等于使用特征的数量，与各个特征的维数无关，因此，其计算复杂度较小，实时性较好。(This code is used to realize covariance tracking in the research field of visual object tracking of computer vision. It enables fusion of various spatial-temporal features into an unified object feature model, hence robust to noise. In addition, its dimension is equal to the number of the features used and independent of the dimensions of the features used. Therefore, it satisfies the requirement of real-time application due to its low computation cost.)

该程序为离焦模糊参数识别的方法,通过canny算子检测边缘，用hough变换检测直线的方法求离焦半径。(The program for the defocus blur parameter identification method by canny operator edge detection using hough transform line detection method for the defocus radius.)

通过高斯核密度估计计算多元变量之间的互信息熵(The mutual information entropy between multivariate variables is calculated by Gaussian kernel density estimation)

利用形态学分水岭算法对图像进行分割，对细胞图像能达到较好的分割效果(Using morphological watershed algorithm for image segmentation, the cell image segmentation can achieve better results)

图像检索，从600多张图片中检索出与目标图片相似度最高的。按照相似度顺序排列。(Image retri , retrieved more than 600 pictures in most similar to the target image. They are arranged in order of similarity.)