小波分析

一个运动估计运算的MATLAB文件，本人已经测试过，非常有用。(A motion estimation MATLAB file operations, I have been tested, very useful.)

infomax ICA的算法实现，并在其中给出了一个infomax ICA的算法用于图像边缘检测的例子(infomax ICA algorithm, which gives an infomax ICA algorithm for image edge detection example)

多目标质心的求取，matlab代码，适合图像处理初学者(Center of mass to strike multiple targets, matlab code for image processing for beginners)

Information of transform watershed

CT图像重建中经典的代数重建算法，其中还有我算的A矩阵的数据(CT image reconstruction in classical algebraic reconstruction algorithm, which there I counted A data matrix)

polsarpro软件的使用说明，适合处理极化SAR图像。(polsarpro the software instructions for dealing with polarization SAR images.)

说明：  通过生成秘钥得到一串具有一定长度的扩频码，将原始消息序列以扩频码的长度过采样产生过采样消息序列，并将扩频码以原始消息序列的长度多次重复成伪随机序列，再令过采样消息序列与伪随机序列异或就可以生成扩频消息序列。在频率域下，将扩频消息序列以一定的嵌入强度自适应地嵌入原始图像，得到的图像具有较强的鲁棒性，比起直接嵌入，扩频水印在受到攻击后恢复出的水印信息误码率要低得多。并且扩频码长度越长，容错率就越高，鲁棒性就越好。(A sequence of spread spectrum codes with a certain length is obtained by generating a secret key. The oversampling message sequence is generated by oversampling the original message sequence with the length of the spread spectrum code, and the spread spectrum code is repeated into a pseudo-random sequence with the length of the original message sequence for many times, and then the spread spectrum message sequence can be generated by making the difference between the oversampling message sequence and the pseudo-random sequence. In the frequency domain, the spread spectrum message sequence is adaptively embedded into the original image with a certain embedding strength, and the resulting image has strong robustness. Compared with direct embedding, the error rate of the watermark information recovered from the attack of spread spectrum watermark is much lower. And the longer the spreading code length, the higher the fault tolerance rate and the better the robustness.)

多目标跟踪算法（3D），针对三维多目标进行跟踪，程序完整，易懂。(The multi-target tracking algorithm (3D), for the three-dimensional multi-target tracking, the program is complete, understandable.)

可以打开视频文件，对视频文件中的人头进行检测并统计数量。算法有HOG RCNN 及Aggrate Channel Features三种(You can open the video file, the head of the video file to detect and statistics. Algorithm HOG RCNN Aggrate Channel Features tionchannel three)

在GDAL库下，采用主成分变换实现多光谱影像与全色高分辨影像的融合。(Under the GDAL library, the principal component transformation is used to fuse multispectral images with panchromatic high resolution images.)