多分辨率下的彩色图像分割方法  

Color Image Segmentation Based on Multi-resolution

作　　者：许利显[1] 潘建寿[1] 唐宏震[1] 

机构地区：[1]西北大学信息科学与技术学院,陕西西安710127

出　　处：《现代电子技术》2008年第12期128-130,共3页Modern Electronics Technique

基　　金：陕西省自然科学基金资助项目(F0306)

摘　　要：通过对彩色图像分割方法的研究,提出一种在多分辨率条件下彩色图像的分割方法。通过离散小波变换获得原始图像的低分辨率近似,并利用分水岭算法对低分辨率近似图像进行初始分割。对初始分割区域根据颜色信息和空间关系按照一定的先后顺序进行合并,直到形成理想的分割结果。对合并后的图像应用小波逆变换得到原始分辨率下的分割结果。在低分辨率下进行的分割和合并大大减小了计算量,提高了分割速度。With the investigation of methods of color image segmentation, a multi - resolution - based new approach is discussed. The Discrete Wavelet Transformation （DWT） is applied to get an approximation of the original image,then the initial segmentation is processed through watershed algorithm. To get a satisfying result,the initial segmented regions are merged in a particular order decided by the color of pixel and connection relationship. Then using the Inverse Discrete Wavelet Transformation （IDWT） to get the result in the original resolution. This approach needs less computation and is much faster than normal methods because the main segmentation and mergence are processed in the low resolution image.

关 键 词：图像分割 小波变换 分水岭 区域合并 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]