欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
图像分割是图像处理到图像分析的关键步骤之一。研究提出了一种基于区域块的聚类分割新算法。高斯混合模型（GMM）聚类算法已广泛应用于图像分割领域，但在真实彩色图像分割中，由于忽略了像素问的空间相关性，使之对高斯噪声非常敏感。首先对彩色图像求其彩色梯度，然后对彩色图像梯度图进行分水岭分割，分水岭分割会产生过分割区域，但基本得到同质区域，提取区域的区域块特征并把其作为高斯混合模型聚类的输入样本值，完成聚类并实现最终分割。新算法把简单的基于像素的聚类提升到基于区域块特征聚类，很好的抑制了噪声对分割结果的影响。通过在合成图像上及大量真实自然彩色图像上进行实验，结果证明本算法能够有效提高分割结果的准确性。
英文摘要：
Image segmentation is one of the key step in image processing and image analysis. A new segmentation algo rithm based on region clustering is presented in this paper. Gaussian mixture model （GMM） algorithm has been widely used in the field of image segmentation. However in the true color image segmentation, the algorithm is very sensitive to Gaussian noise because of the ignorance of spatial correlations between pixels. In the paper, firstly the color image gra dients are calculated. Then, the watershed segmentation on the color image gradient is implemented. Watershed segmenta tion produces oversegmentation area, but the homogeneous regions are basically obtained. The characteristics of the re gional blocks are then calculated and taken as the input sample values of the Gaussian mixture model clustering to com plete the clustering and achieve the final segmentation. The impact of noise on the segmentation result can be decreased by using the characteristics of regional blocks instead of pixels in the new clustering algorithm. The segmentation experi ment results of synthetic color images and lots of nature color images prove that the accuracy of segmentation used the algorithm in this paper can improve effectively.
同期刊论文项目
同项目期刊论文
Multi-software reliability allocation in multimedia systems with budget constraints using Dempster-S
期刊信息
位置：