IMM-algorithm-based-target-tracking

采用卡尔曼滤波算法进行视频序列图像的跟踪，这个程序从简单易懂的角度实现了视频目标的跟踪。(Kalman filtering algorithm for video sequence image tracking, this process from the perspective of straightforward realization of the video target tracking.)

说明：  dctcom.m文件利用DCT变换完成对输入图像进行压缩；imagecbe.m完成对输入的两幅RGB图像用小波分析的方法进行图像融合 imagecom.m完成对输入的RGB图像用小波分析的方法进行自动降噪，得到高频系数阈值，降噪效果百分比和结果 wavelet1D.m完成对输入的一维信号进行多尺度离散小波分解 wavelet2D完成对输入的二维信号进行多尺度离散小波分解；zigzag.m完成对输入的8*8矩阵按照zigzag排列抽取数据.(document the use of DCT transform dctcom.m completion of the input image compression imagecbe.m completion of the two RGB input image with wavelet analysis method of image fusion imagecom.m completion of the input RGB image using wavelet analysis method automatically drop noise, the threshold of high-frequency coefficient, noise reduction effect and the results of the percentage of wavelet1D.m completion of the input signals of one-dimensional discrete wavelet multi-scale decomposition wavelet2D completion of the input signals of two-dimensional discrete wavelet multi-scale decomposition zigzag.m completed the input of 8* 8 matrix of data collected in accordance with the zigzag order.)

提出了一种结合USFFT Curvelet 变换的各向异性扩散图像去噪模型。它有机结合了Curvelet 变换和各向 异性扩散（P-M扩散）两者的优点。通过P-范数方法选择合适的梯度阈值K，P-M扩散过程通过处理经过Curvelet 变换得到的图像的不同尺度的Curvelet 系数矩阵，实现了建立在对图像多尺度分析的基础上的新P-M扩散模型。 实验表明，新模型的处理结果能有效避免传统P-M 扩散出现的阶梯效应，同时更好地保留图像的纹理和细节。 关键词：图像去噪算法；各向异性扩散；Curvelet 变换；P-M 扩散(Abstract: An image de-noising model by integrating anisotropic diffusion with USFFT Curvelet transform was proposed, which combined the strongpoint of Curvelet transform with anisotropic diffusion (P-M diffusion). By choosing appropriate gradient threshold K through p-norms and carrying out the P-M diffusion process depend on the different scale matrixes of Curvelet coefficient of the image from Curvelet transform iterations, as a result, the improved model made it possible to carry out the new P-M diffusion de-noising process based on multi-scale analysis of the image. The experiment results have demonstrated that the model can avert the stair-casing effect in the traditional P-M diffusion effectively and keep the textures and details of images better.)

用ＷＰＦ写一个简单的转盘

3d 重建 包括特征点的提取，匹配，以及三维点云的建立(3d reconstruction ,including feature detection and 3 D points cloud ,the change from 2D to 3D)

说明：  一个双目匹配求解视差图的matlab源代码(A binocular matching to solve the disparity map of Matlab source code)

将一幅M×N的灰度图像用3×3平均滤波器进行模糊，分别再加上一定的高斯 噪声和均匀噪声。然后，设计一个维纳滤波器对这两幅图像进行复原，分别 计算这两幅图像复原前后的PSNR。(Gray-scale image of a MN 33 average filter is vague, respectively, plus a Gaussian noise and uniform noise. Then, design a Wiener filter to recover these two images, the two images before and after the recovery of PSNR were calculated.)

道路识别的一个matlab仿真程序，主要是基于道路颜色的图像处理。(A road recognition matlab simulation program was based primarily on the road color image processing.)

图像几何校正算法。采用分段校正方法对图像进行校正。(Image geometric correction algorithm. The use of sub-correction method of image correction.)

此算法为基于k-means聚类的非局部均值快速去噪算法，方法便捷有效，可保持边缘信息(this is a fast edge-preserving denoising method based on non-local means method with k-means clustering)