nmfsc

通过阈值去噪，最关键的就是要确定最佳阈值T 的大小.如果阈值太小，可以保留更多的小波系数，但同时噪声也被保留；相反，阈值太大，可以消除更多的噪声，但重要图像特征又将被滤掉，引起偏差.目前使用的阈值分为全局阈值和局部阈值.全局阈值对各层所有的小波系数或同一层内的小波系数都是统一的；局部阈值则是根据当前系数周围的局部情况来确定合适的阈值，具有灵活性.(Threshold denoising, the most critical is to determine the optimal threshold T s size. If the threshold is too small, you can retain more of the wavelet coefficients, but the noise was retained the contrary, the threshold is too large, can be eliminatedmore noise, but the important image features in turn be filtered out, causing bias current threshold is divided into a global threshold and local threshold. global Threshold the wavelet coefficients of all the wavelet coefficients of the layers or on the same floorare unity local threshold is based on the local situation of the current coefficient around to determine the appropriate threshold, with flexibility.)

说明：  可自选输入和参数的中值/均值/拉伸滤波，各项参数说明详见.m文件注释(Median/Average/Linear Stretch Filter)

这是对图像进行分块的代码，分块后提取LBP特征，站中有LBP代码，可以直接调用这个代码进行分块LBP特征提取(This is to block of code of image, block LBP feature extracting, after standing in LBP code, you can directly invoke the code block LBP feature extraction)

sift角点检测及匹配，在不同尺度空间下的角点检测方法，很方便实用。(sift corner detection and matching, the corner detection methods under different scales of space, it is convenient and practical.)

稀疏矩阵字典的优化KSVD，本人已进行测试，效果甚佳,可以用在图像，数据的压缩和分类(Sparse dictionary optimization KSVD, I have been tested, very good results)

根据RBM深度网络（Hinton，2006）进行MINST手写文字的分类器的训练，利用训练得到的权值制作了这个小程序，通过这个程序可以看出训练结果对数据集内的测试样本和训练样本都能进行很好的识别，但是对其他的手写字体识别就没有那么好了。(According to RBM depth network (Hinton, 2006) conducted MINST handwritten text classifier training, using the training to get the right value to produce this small program, this program can be seen through the results of the training data set and the training sample test samples are can be well identified, but on the other handwriting recognition is not so good.)

本程序用量子行为的粒子群算法训练支持向量机，并用IRIS数据验证了该方法的有效性(This procedure acts with quantum particle swarm optimization training support vector machines, and IRIS data to verify the effectiveness of the method)

快速TV算法用于图像去模糊、图像去噪等，其中TV模式包括两种(TV went fast algorithm for image blur, image denoising, which includes two TV mode)

内容包括高斯混合概率假设密度滤波器仿真程序，GM-PHD滤波器提供了一种PHD滤波器工程化的方法，在无数据关联的情况下进行目标跟踪。(This algorithm is extended to accommodate mildly nonlinear target dynamics using approximation strategies from the extended and unscented Kalman filters.)

实现多层tiff显微图像识别，追踪显微镜下PS微球的坐标并给出颗粒运动轨迹(Multi-layer TIFF micro-image recognition is realized. The coordinates of PS microspheres under the microscope are tracked and the trajectories of particles are given.)