一种新的基于区域合并的图像分割算法

提出了一种新的彩色图像分割算法,以区域合并为基础并使用联合彩色纹理直方图.利用彩色直方图与最大相似性的区域合并算法已成功地应用在彩色图像分割,在某些情况下,只使用彩色直方图信息并不能充分且有效地达到优越的分割.所提出的方法将彩色直方图和纹理直方图的信息用于测量不同地区的相似性,从而引导区域合并的进程.此外,为了获取纹理信息,采用局部二进制模式(LBP),以便嵌入主要的统一LBP模式来测量不同地区的相似性.实验结果表明,与基于经典图像分割方法的区域合并相比,该算法取得了很大的进步.此外,与其他类似的方法相比,该算法更加健全和精确.

Study on A New Image Segmentation Method Based on Region Merging

This paper proposes a novel color image segmentation method based on region merging by using joint color-texture histogram. The region merging algorithm by maximal similarity with color histogram has been applied successfully in color image segmentation, contrastively, in some cases, only using color histogram information is not sufficient and effective for superior segmentation. The proposed method incorporates both color histogram and texture histogram information to measure the similarity of different regions and thus to guide the region merging process. Moreover,＇ in order to capture the texture information, our method adopts the local binary pattern （LBP） frames so that the major uniform LBP patterns are embedded to measure the similarity of different regions. Experiments show that a great improvement has been made compared with region merging based methods of classic image segmentation. Also, the proposed algorithm is robust and accurate in comparison with other similar methods.