TV图像处理

图像重构实现干涉条纹缺陷图的缺陷检测，能很好的滤除条纹，得到缺陷(Interference fringe image reconstruction to achieve defect detection defect map can be a very good filter stripes)

一种基于节点密度分割和标签传播的Web页面挖掘方法(A mining based on node density segmentation and labeling dissemination of Web pages)

自动前背景分割，基于超像素、形状和外观前验概率及图切割(A novel foreground/background segmentation algorithm that attempts to segment the interesting objects from the rest of the image.The model includes a geometric prior and an appearance prior, whose parameters are learnt on the fly from images that are similar to the query image. It uses graphcut based energy minimization to enforce spatial coherence on the model’s output.)

一种像素级指纹方向场估计算法以及对应的像素级奇异点检测算法。(A pixel-level orentation field estimation algorithm and the corresponding singular point detection method.)

DCT域图像水印算法，并根据携密载体提取秘密信息(DCT domain image watermarking algorithm, and extract secret information according to the carrier.)

用于SIFT变化的关键点判断，定对比度和herrian矩阵，最后结果为关键点在DOG里的坐标！(Change for the SIFT key points to determine, set contrast and herrian matrix, the final results for the key points in the coordinates in the DOG!)

说明：  简单的介绍了基于matlab的QR码的识别，包括图像的预处理，图像定位，图像的识别等(This paper briefly introduces the recognition of QR code based on matlab, including image preprocessing, image positioning, image recognition and so on.)

这是一种简单的互相关图像匹配算法，已经在matlab环境下测试过，效果很好。附带两张匹配图片(This is a simple cross-correlation image matching algorithm has been tested in the matlab environment, the effect is very good. With two matching pictures)

高光谱图像MNF算法，实现高光谱图像的去噪，注释详尽(Hyperspectral image MNF algorithm to achieve denoising hyperspectral images, detailed comments)

基于PCNN的特征提取,PCNN用于特征提取时，具体平移、旋转、尺度、扭曲等不变性，这正是许多年来基于内容的图像检索系统追求的目标，同时PCNN用于特征提取时，有很好的抗噪性。而且PCNN直接来自于哺乳动物视觉皮层神经的研究，具有提取图像形状，纹理，边缘的属性。用PCNN能很好地对图像进行签名，将二维的图像的特征提取成一维矢量签名。(Feature extraction of specified object is an important preprocessing stage in machine vision systems. In this paper, we present a novel hybrid feature extraction method using PCNN (Pulse Coupled Neural Network) and shape information. First, we use PCNN firing map train to formulate object’s time signature, then we use roundness of each firing map to formulate object’s shape information vector, the final feature matrix we got is combined time signature and roundness. We take correlations as our judge criteria in our experiments. It has been proved that the algorithm is not sensitivity with the rotation, scaling and translation of the object and is a useful method for target recognition applications.)