datamining

DB97提升小波程序，实现任意偶数大小图像双正交97提升小波变换，比传统的CDF97卷积算法好(DB97 lifting wavelet procedures, arbitrary size images even biorthogonal wavelet transform 97 upgrade, than the traditional filter convolution algorithm good)

首先利用形态学开运算将图像中小于结构元素的亮点去掉，利用源图像与处理后的图像作差，对差图像进行阈值分割，得到二值图像，将二值图像中的一些干扰区域去掉，提取二值图像中的圆形区域，并在源图像上对小目标进行标记，实现对圆形亮目标的检测(Remove the area in the image smaller than structural element by morphologic open operation，receive image A，original image subtract image A，threshold segmentation for difference image，receive binary image，remove the interference region in the binary image，then extract circular area，and sign the target in the original image. Realize the circular bright target detection)

链码直线检测，采用求取图像边界，然后边界跟踪，判别直线(Chain code line detection, to strike the image boundary, then the boundary tracking, discriminant linear)

在视频文字检测以及提取方面，比较好的参考论文，大家可以下载参考！(Video text detection and extraction, a good reference papers, everyone can download the reference!)

使用VC++ 6.0 和 OpenCv实现粒子滤波目标跟踪算法，主要功能在MFC对话框内实现。(particle filter object tracking)

本程序完成了对运动图像的识别与跟踪功能，程序里自带了测试的图像(The program to complete the identification and tracking of moving images, the program comes with the test image)

1.设计一个程序，对一幅灰度图像的实现如下几何变换,1)放大为原图1.5倍 2)绕中心旋转30度(CW) 3)采用偏移量插值实现一个透视变换. 灰度差值用最近邻插值和双线性插值 2. 绘制一幅灰度图像的梯度幅度图像（三点法求梯度），针对梯度幅度图像合理的选择一个阈值（通过试验即可）将其二值化，以获得图像边缘检测图像。 ((1) design a program, a grayscale image as geometric transformation, zoom 1) for 1.5 times the original 2) around the center of rotation of 30 degrees (CW) 3) interpolation of the offset to achieve a perspective transformation.gray level difference between the nearest neighbor interpolation and bilinear interpolation (2) draw a grayscale image of the gradient magnitude image (three-point method seeking gradient), the gradient magnitude image is reasonable to select a threshold value (the test can be) the second value to the image edge detection image.)

VC++,DirectX,运动图像检测,运动轨迹跟踪(VC++, DirectX movement trajectory tracking image detection)

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.)

完成对视频帧的四叉树分割，在纹理复杂处分割块较小，在纹理平坦处分割较大(Complete quadtree segmentation of video frames, complex area in texture segmentation smaller, flat larger regional segmentation block in texture)