QQ客服 官方微博 反馈留言

	基于小波变换的目标边缘搜索分割方法
	李国嵩 ; 孟卫华 ; LI Guo-song ; MENG Wei-hua
	2010-06-09 ; 2010-06-09
关键词	边缘检测 反双正交小波 梯度算子 启发式搜索 目标识别 Edge detection Reverse biorthogonal wavelet Gradient operator Heuristic search Target recognition TP391.41
其他题名	Target edge searching segmentation method based on wavelet transform
中文摘要	为满足目标识别实时性、抗噪声的要求,提出了利用小波变换的图像边缘搜索分割方法。用反对称双正交小波算子对图像进行小波变换,对小波变换模极大值进行启发式搜索检测边缘,得到目标边缘的链码表示。该搜索方法能较好地克服噪声干扰,边缘丢失时能通过在模值分布图中搜索找回边缘,增强算法的鲁棒性。实验结果表明,在同样平台下,使用该方法检测边缘能减少一般小波多尺度分析的计算量,计算速度与使用Sobel算子的梯度方法相当,且具有更好的抗噪声能力。该方法边缘连接良好率高,特征提取方便,综合提高了图像处理算法效率。; In order to meet the requirement of real-time and noise immune capability for target recognition,the target edge searching segmentation method based on wavelet transform was presented. The reverse biorthogonal wavelet was used, and then the target edge was detected using a heuristic search in modulus maxima of the wavelet transform. The target edges and a chain code representation of the segmentation result were obtained along with the search at last. This searching method was able to achieve better anti-noise performance in edge detecting. Meanwhile the method could find the edges lost in searching by seeking in modulus distribution and this would highly improve the robustness of the algorithm. Demonstrations show that the proposed algorithm avoids the large calculation in common multi-resolution wavelet analysis under same testing condition. Compared with the gradient algorithm using Sobel operator, the algorithm has almost the same computing speed but much better anti-noise ability. The result of connection rate of complete edge is high and can make the feature detection part more convenient. And then efficiency of the image processing algorithm was improved synthetically.; 国家预研基金资助项目
语种	中文 ; 中文
内容类型	期刊论文
源URL	[http://hdl.handle.net/123456789/54579]
专题	清华大学
推荐引用方式 GB/T 7714	李国嵩,孟卫华,LI Guo-song,等. 基于小波变换的目标边缘搜索分割方法[J],2010, 2010.
APA	李国嵩,孟卫华,LI Guo-song,&MENG Wei-hua.(2010).基于小波变换的目标边缘搜索分割方法..
MLA	李国嵩,et al."基于小波变换的目标边缘搜索分割方法".(2010).

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。