First , the source image was decomposed by wavelet lifting , using mathematical morphological method extracts edges in low - frequency approximate image and using wavelet transform method extracts edges in high - frequency detailed images on the different levels , then the two edge images were fused according to some fusion rules to obtain an integrated and clear edge image 对源图像进行小波分解，用数学形态学法对低频子图像进行边缘检测，用小波变换法对不同分解层上的高频子图像进行边缘检测，采用一定的融合规则将这两个边缘图像融合在一起得到一幅完好的边缘图像。
The writer analyzed mallat three criteria and introduced the optimized function based on these three criteria , at last found optimized mother wavelet for edge detection . at the end of this chapter single edge searching method based on local maximum is introduced and gained ideal edge image 分析mallat提出的评价边缘检测函数优劣的三个判别函数，利用这三个判别函数列出最优化方程，推导出边缘检测最优小波，最后创造性提出局部最大值单像素边缘搜索法，提取出单像素宽度边缘图像，图像效果达到了满意的效果。
In this paper , we define quadratic spline as wavelet , do the two dimentional dyadic wavelet transform on cell image , and get local modulus maxima from wavelet transform ' s results - modulus and angles , so we can find the cell image ' s edge image in each scales , at last , we compute optimum scale of cell image edge detection , and receive a good edge image which synthesize the characters in each scale 本文中我们用二次样条小波作为小波基函数，对细胞图像进行二维二进小波变换，计算小波变换结果的局部模极大值点，得到各个尺度下的细胞图像的边缘，计算细胞图像边缘检测的最优尺度，最后得到综合了各个尺度特征的较好的细胞图像的边缘。
Edge usually owns the main information of image and lies in the anomalous and unstable location which gives the position of image profile . the profile is usually the most important characters of image during the processing of image , so it is necessary for us to detect the image and get the edge image 边缘往往携带着一幅图像的大部分信息，而边缘存在于图像的不规则结构和不平稳现象中，也即存在于信号的奇异点或突变处，这些点给出了图像轮廓的位置，这些轮廓则常常是在图像处理时所需要的非常重要的一些特征条件，这就需要对一幅图像检测和提取出它的边缘。
This paper proposes a new algorithm , called edge - based texture driven shape model e - tdsm , for non - frontal face alignment task . first , the texture is defined as the un - warped edge image contained in the shape rectangle ; then , a bayesian network is constructed to describe the relationship between the shape and texture models ; finally , expectation - maximization approach is utilized to infer the optimal texture and position parameters from the observed shape and texture information 姿态的变化使得人脸形状的分布变得非常复杂，特别是侧脸图像，由于自遮挡的原因，采用单一的模型有时候很难表达人脸形状的分布同时，再加上光照的变化，一方面使得人脸轮廓容易和背景混淆，另一方面有可能会在人脸上形成伪边界，使得轮廓线被错误地定位到这些伪边界线上。
The developed simple m - s model is suitable for weak edge image segmentation by adjusting weighting coefficient . the experiments on tumor ct image segmentation show that geometric active contour model requires more computational load than snake model but it can converiently control the topology construct of segmented image 在肿瘤ct医学图像分割实验中,通过比较参数活动轮廓模型与几何活动轮廓模型可知,参数活动轮廓模型计算量小于几何活动轮廓模型计算量,而几何活动轮廓模能适应拓扑结构的变化。
The edge of image is often considered as its fundamental feature in image processing . it has been used in image processing and analyzing technologies such as feature description , image recognition , image segmentation , image enhancement and image data compression on a high level . then we can comprehend and analyze the edge image later on 在实际的图像处理问题中，图像的边缘图作为图像的一种基本特征，被经常应用到较高层次的特征描述、图像识别、图像分割、图像增强以及图像压缩等等的图像处理和分析技术中，从而可对图像作进一步的分析和理解。
The edge of image is often considered as its fundamental feature in image processing . it has been used in image processing and analyzing technologies such as feature description , image recognition , image segmentation , image enhancement and image data compression on a high level . then we can comprehend and analyze the edge image later on 在实际的图像处理问题中，图像的边缘作为图像的一种基本特征，被经常应用到较高层次的特征描述、图像识别、图像分割、图像增强以及图像压缩等的图像处理和分析技术中，从而可对图像作进一步的分析和理解。
In the shape - based image retrieval , transforming the image with wavelet modulus maximum to get multi - scale edge images , then employs a set of seven invariant moments to extract the features of image , which can effectively resolve the variance with respect to translation , scaling , and rotation 在基于形状的图像检索中，采用小波变换模极大值的方法来提取图像的多尺度目标边缘信息，然后采用多尺度的不变矩作为特征来图像相似性检索，可以有效的解决因图像的平移、尺度、旋转变换等带来的问题。
But this way is not considering the information of texture and not doing more works . so this edge image exists much flaw , such as , disorder interfereing points , fake edges , discontinuity in the edges , etc . finally , this thesis considered the traditional way not to adapt to our problem for developing the 3 - d stimulating of athlete 这样做既没有考虑到原始图像序列中所附带的纹理信息，也没有对已检测得到的轮廓进行深入的、细致的处理，从而使得轮廓图像中存在很多的干扰信息，比如：尚未处理的杂乱孤立干扰点、伪边缘、运动目标边缘不连续等。
