inexact_alm_rpca

读取RAW格式文件，将其转换为BMP格式文件并存储，主要实现灰度图像的转换(Read the RAW format, convert it to BMP format file and stored, mainly to achieve the conversion of grayscale images)

Itti模型的显著图提取，使用matlab编写，很好的代码，可供参考学习(Itti model saliency map extraction, the use of matlab prepared very well the code)

标准Hough检测圆，在数字图像中检测圆形目标，需要先边缘提取，在二值图像中检测圆。(circle detection using standard hough transform )

基于Retinex的MSRCR图像增强算法的源代码，已经测试，可以直接使用。(Based on the Retinex MSRCR image enhancement algorithm of source code, have been tested, can use directly )

基于DCT域的数字水印算法主要包括水印调制、水印嵌入和水印提取，首先在嵌入载体图像之前用Logistic混沌序列对水印进行调制，得到一个只有+1和-1随机序列的水印，同时将载体图像进行全局DCT变换，然后将水印图像嵌入人类视觉系统最重要的部分，即DCT最大的系数，DC（Direct Current）部分也包含在这最大的系数中，这部分的系数满足人类视觉特性，会使水印嵌入的性能更好。嵌入算法是将水印与这些系数进行相乘获得新的系数，再用新的系数来替换原来的系数，即获得嵌入水印后的离散余弦域的图像，然后对这个变换后的图像进行IDCT，就会得到需要的水印图像。(While the carrier image digital watermarking algorithm based on DCT domain include watermark modulation, watermark embedding and watermark extraction, the first of the watermark embedded in a carrier modulated image before using Logistic chaotic sequence, only+1 and-1 to get a random sequence of watermark, Global DCT transform, then the watermark image is embedded most important part of the human visual system, i.e. the largest DCT coefficients, DC (Direct Current) in this section also contains the largest coefficient, the coefficient of this part of the human visual characteristics to meet, make watermarks Embed better performance. The watermark embedding algorithm is with these coefficients are multiplied to obtain new coefficients, and then a new factor to replace the original coefficients, namely discrete cosine domain to obtain an image watermarked and then transformed the image of this conduct IDCT, on watermark image will be needed.)

深度学习相关代码实现，可以用来做去噪，去模糊，回归，识别等( Initialize and train an RBM energy-based model. Supports binary, Gaussian, and multinomial inputs (still in development). Supports binary and noisy rectified linear units (NReLU) in the hidden layer. Supports joint modeling of binary and multinomial variables for classification.)

利用matlab对多光谱遥感影像进行水体指数（ndwi）以及植被指数（ndvi）提取(Extraction of Water body Index (ndwi) and vegetation Index (ndvi) from Multispectral remote Sensing Images by matlab)

meanshift算法的车辆跟踪，效果非常好，改进版的，希望对大家有用(Meanshift algorithm of vehicle tracking, the effect is very good, improved version of the hope of the useful)

基于sift的图像特征点提取和图像拼接，有可直接运行的代码和图片(Sift the image feature point extraction and image stitching, can be directly run the code and pictures)

Demo for smoke effects on old NVIDIA GPUs.