ZhifangtuJunheng

轻量级的MTCNN实现，无需caffe和tensorflow基本库支持，内部实现卷积运算，速度比普通mtcnn要快一倍。

包括计算共生矩阵、对共生矩阵归一化、对共生矩阵计算能量、熵、惯性矩、相关4个纹理参数、求能量、熵、惯性矩、相关的均值和标准差作为最终8维纹理特征 (Including the calculation of co-occurrence matrix, normalized to the co-occurrence matrix, co-occurrence matrix calculation on the energy, entropy, moment of inertia, related to four texture parameters, find the energy, entropy, moment of inertia, related to the mean and standard deviation as the final eight-dimensional texture feature)

说明：  对给定图片求图片中的染色体个数。具体步骤：中值滤波方法，二值化处理，膨胀和腐蚀，开闭运算，求连通区的个数(Of a given picture for the picture of the chromosome number. Concrete steps: median filter method, binarization processing, expansion and corrosion, opening and closing operations, and the number of connected areas)

利用Matlab编写的基于hough变换的椭圆检测程序，已经运行过的，附图(Written in Matlab, hough transform based ellipse detection procedure has been run off, with photos)

几种全参考图像质量评价标准，例如VIF、IFC、VSNR、PSNR、MSE等。(Several full reference image quality assessment standard, such as 2 VIF is based, the IFC, VSNR, PSNR, MSE and so on. )

sift匹配算法 尺度不变变换 特征描述子 特征点匹配 (sift match)

图像融合拼接GUI设计，包括图像融合和图像拼接的matlab实现。(Image fusion splicing GUI design, including image fusion and image stitching matlab implementation.)

采集一幅像素大于64*64黑白图像； 常规的数学统计数据处理：计算图象各象素点灰度值得均值、标准差、方差，并绘出灰度直方图； 采用[FFT（傅立叶变换）]对图像进行分析(Acquisition of a pixel is greater than 64* 64 black and white images The conventional mathematical statistical data processing: calculation of each pixel grayscale image is worth the mean, standard deviation, variance, and draw grayscale histogram The [FFT (Fourier transform)] to do image analysis )

一种新的目标跟踪算法。将压缩感知引入到序列图像目标跟踪中，可以实现目标的稳定快速跟踪。(A new target tracking algorithm. The compressed sensing is introduced into the sequence of image target tracking, stabilization target can be achieved fast track.)

脑电信号 分析 fast ica去噪，去除心电，眼电，肌电，效果不错(fast ica for eeg)