computer_vision_calibr

图像的各向异性扩散，利用matlab实现Perona_Malik算法，梯度阈值K通过对图像求解2-范数的方法迭代获得(Anisotropic diffusion images, using matlab to achieve Perona_Malik algorithm, the gradient threshold images by solving the K 2- norm method will be obtained)

说明：  基于L1范数的多帧图像超分辨率图像重建算法，在原有算法基础上改进，提高重建精度和效率。(Based on the L1 norm of the multi-frame image super-resolution image reconstruction algorithm, the improvement on the basis of the original algorithm to improve the reconstruction accuracy and efficiency.)

图像金字塔分解，输入一幅图像，以及需要分解的层数，可以输出图像的金字塔分解(Image pyramid decomposition)

基于霍夫变换开发一种较通用的快速圆检测方法，并将之应用到虹膜图象的检测中，完成对虹膜内、外边缘的定位。(Hough transform based on the development of a more general-purpose fast circle detection methods, and to apply it to the detection of iris images, the completion of the iris inner and outer edge of the positioning.)

raw 数据转换成plant 的yuv 数据，已经测试通过。希望对raw数据进一步处理的同学有帮助。( Raw data is converted into YUV plant data, has been tested through. Hope that the raw data to further deal with the students to help. )

图像颜色特征的提取：HSV颜色直方图 描述的是图像的整体颜色情况。本程序采用非均匀量化方法得到36bin的颜色直方图。(Image Color Feature Extraction: HSV color histogram describes the situation of the image' s overall color. This procedure using non-uniform quantization method the color histogram 36bin.)

说明：  使用Harris角点检测器查找特征。再使用SIFT描述符并匹配两个图像之间的特征描述符。使用RANSAC执行单应。最后把图像包起来缝合。(Use Harris corner detector to find features. Sift descriptors are then used to match the feature descriptors between the two images. Use RANSAC to perform single response. Finally, the image is wrapped and stitched.)

指纹识别 能够快速的识别出指纹库中的指纹，是否与输入的待检测的指纹相同，效果很好(Fingerprint identification can quickly identify the fingerprint of fingerprints, whether the same as the input fingerprint to be detected, the effect is very good)

在图像处理过程中盲解卷积也是一个很好地图像去噪的方法。(In the image processing method of blind deconvolution is also a good image denoising.)

首先生成如下形式的图片：图中的白条为7象素宽，210象素高，白条间的宽度为17象素。 应用多种滤波器进行处理：算术均值，几何均值，谐波均值，逆谐波均值，中值滤波，中点滤波，最大值、最小值滤波等。注意，需改变模板大小。详细比较上述的结果，并分析原因。 (First generation following form images: figure IOUs 7 pixels wide, 210 pixels high, the width is 17 pixels between IOUs. Using a variety of filters for processing: the arithmetic mean, geometric mean, harmonic mean, inverse harmonic mean, median filtering, filtering midpoint, maximum, minimum filtering. Note that the need to change the template size. A detailed comparison of these results and analyze the reasons.)