mjfap

说明：  自己实现的BM3D图像去噪算法。它首先把图像分成一定大小的块，根据图像块之间的相似性，把具有相似结构的二维图像块组合在一起形成三维数组，然后用联合滤波的方法对这些三维数组进行处理，最后，通过逆变换，把处理后的结果返回到原图像中，从而得到去噪后的图像。(Image Denosing by Sparse 3-D Transform-Domain Collaborative Filtering.)

说明：  调用标定好的双目摄像机，利用其内外参数进行立体匹配，计算所拍图像的视差，产生深度图(Call the calibrated binocular camera, use its internal and external parameters for stereo matching, calculate the parallax of the captured image, and generate depth map)

灰度共生矩阵和灰度梯度共生矩阵纹理提取，能够计算纹理的多个测度(Gray Level Co-occurrence matrix and gray-gradient co-occurrence matrix texture extraction)

说明：  实现j基于patch的图像修复，可以比较好的实现功能(The image repair based on Patch can realize better function, based on Matlab not relayed on C.)

DNA比对中FASTA算法的简单实现，通过建立查找表、位移表和位移向量表来得出两条序列的最大匹配位移。(DNA matching the FASTA algorithm is simple to achieve, through the establishment of a lookup table, displacement and the displacement vector table to derive maximum matching two sequences displacement.)

说明：  此jar包用于解析jbig格式文件的工具包，调用接口即可(This jar package is a toolkit for parsing JBIG format files and calling interfaces.)

说明：  点云插值，离散点云插值，孔洞填充，使用python语言(Point cloud interpolation, discrete point cloud interpolation, hole filling, using Python language)

DensePose用深度学习把2D图像坐标映射到3D人体表面上，再加上以每秒多帧的速度处理密集坐标，最后实现动态人物的精确定位和姿态估计。该技术集目标检测、姿态估计、目标部分/实例分割等多种计算机视觉任务于一身的一个综合问题。(DensePost maps 2D image coordinates to 3D human body surface by in-depth learning, and processes dense coordinates at the speed of multiple frames per second. Finally, it realizes precise positioning and attitude estimation of dynamic characters. This technology integrates many kinds of computer vision tasks, such as target detection, attitude estimation, target part/instance segmentation, etc.)

LVQ(学习矢量量化)算法：它是Kohonen的有监督学习的扩展形式,融合了自组织和有导师监督的技术,学习方法是竞争的,但(LVQ (Learning Vector quantization) algorithm: it is an extended form of supervised learning of Kohonen, which combines self-organizing and supervised techniques. The learning method is competitive, but the)

文字可视化，对于给定的文本，输出可视化效果的程序。(word visualization)