peizhun

1阶HMM模型多个观察值序列的评估和训练问题模型。适合用于图像纹理的去噪，分割等问题。程序用于为网友提供学习之用，由于是自编自用，所以如有不通的地方仅供参考。另由于数据较大，没有上传。(Evaluation and Training of multiObservation sequence Used by 1-order HMM)

利用多尺度高斯函数去除场景的光照分量；然后构造了一种二维伽马函数，利用光照分量的分布特性调整二维伽马函数的参数，降低光照过强区域的亮度值，提高光照过暗区域的亮度值，最终实现对光照不均匀图像的自适应校正处理。(The multi scale Gauss function is used to remove the illumination component of the scene, and then a two-dimensional gamma function is constructed. The parameters of the two-dimensional gamma function are adjusted by the distribution characteristics of the illumination component, the brightness value of the light illuminated region is reduced and the brightness value of the light dark area is improved. Finally, the adaptive correction of the uneven illumination image is realized. Positive treatment.)

利用混合高斯模型对图像序列经行背景建模，并保存。（附图片）(Using Gaussian mixture model for background modeling image sequences through the line and save. (With pictures) )

强大的小波包分析程序，将信号分频分析，在感兴趣的频率域内抽取信号信息。(Powerful wavelet packet analysis procedures, the signal of frequency analysis, frequency domain of interest extracted signal information.)

一个改进的卡尔曼滤波器 用来进行运动目标的跟踪 (An improved Kalman filter used for tracking moving targets)

说明：  克里金算法2.0版本，对二维插值有详细的工具箱(kriging algorithm V2.0 for matlab )

根据图像的边界进行轮廓跟踪来获取图像的外轮廓线(The image of the contour tracking to obtain an image of the outer boundary contour)

图像分解算法，四种方法可选，包括threshold, K means threshold, region growing以及gradient based segmentation(Image decomposition algorithm, four methods available, including the threshold, K means threshold, region growing and gradient based segmentation)

主要是关于模糊熵的，里面还和灰色关联度有点联系。搞图像的应该可以看的懂的。(Mainly on fuzzy entropy, which has little contact and the gray. Should be able to read out the image of understand.)

sift匹配与识别演示程序 运行face.m文件，载入图片后可提取sift特征点(sift demonstration.rar. run face.m)