graphcuts

光栅图形学中的扫描填充算法。list<Line> AET和list<Line>* NET构成了活性边表和新边表.算法先初始化各扫描线新边表,之后对于每条扫描线更新活性边表，并且配对交点来着色。 (Raster graphics scanning fill algorithm. list <Line> AET list <Line>* NET constitute the active side of the table and side table new algorithm to initialize each scan line side table new, after each scan line update the active side of the table, and the the paired intersection of coloring.)

JPEG压缩方法由于其较高的压缩比和理想的压缩效果，是目前应用最广泛的图像压缩方法(JPEG compression method because of its higher compression ratio and ideal compression effect, is currently the most widely used method of image compression)

说明：  使用matlab生成动态的心源程序，可自己修改，表白女神，好东西要分享(Use matlab to generate a dynamic heart to express the goddess and share good things)

包括回归分析和概率统计，结合PCA的尺度不变特征变换(SIFT)算法，计算晶粒的生长，入门级别程序。( Including regression analysis and probability and statistics, Combined with PCA scale invariant feature transform (SIFT) algorithm, Calculation of growth, entry-level program grain.)

基于轨迹划分的异常轨迹检测方法。首先采用两种分段粒度将每条轨迹分为若干子轨迹线段，然后，在基于距离的异常检测方法的基础上，引入调整系数的概念，采用密度与距离相结合的方法进行异常轨迹检测。 (Abnormal trajectory detection method based on trajectory partition.Firstly each trajectories should be divided into several substrajectories using two granularities. Then the notion of adjustment coefficient is introduced on the basis of distance based anomaly detection method, and the abnormal trajectories are detected combining density and distance. )

计算特征的汉明距离作为特征匹配相似度的评价标准(The Hamming distance of the feature is used as the evaluation criterion of feature matching similarity)

基于人群分类的SEIR模型的仿真，得出模型中各参数(传播率，等)对谣言传播范围的影响(Simulation of SEIR model based on crowd classification)

说明：  对三角划分算法的研究，对以前方法的改进，以达到良好的效果，(Triangulation algorithm to study the improvement of previous methods in order to achieve good results,)

说明：  基于CV模型的自动分割,能够对灰度图分割操作(image segmentation Level set Chan-Vase)

Cohen类时频方法采用不同的核函数来抑制交叉项，本质上是在模糊域对信号进行低通滤波，其中典型的方法有Choi-Williams 分布和Cone-shaped 分布等。(The Cohen time frequency method uses different kernel functions to suppress the cross term, which is essentially a low pass filter for the signal in the fuzzy domain. The typical methods are Choi-Williams distribution and Cone-shaped distribution.)