摘要
纹理分割是图像处理的难点之一。针对此问题,提出了一种基于局部二进制模式(localbinary pattern,LBP)驱动的区域围道分割模型。该模型首先将均匀模式的思想用于LBP/C算子,使纹理模式的数量减少了77%,明显降低了提取纹理特征所需的时间;其次对无边缘活动围道模型进行了改进,使其能用纹理特征来演化曲线或曲面分割纹理图像;然后用多级分层的策略对提出的模型进行了延拓,可用于分割多类目标的图像,避免了多相位模型初始围道难以选择的问题,提高了模型收敛的速度;最后运用AOS(additive operator splitting)算法以改善模型求解的效率,进一步提高了图像分割的速度。对合成纹理图像和遥感图像的实验结果说明,提出的分割方法具有分割速度快、精度高的优点。
A new method is proposed for texture image segmentation. In the proposed method, we apply uniform LBP/C （local binary patterns and pattern contrasts） operator to extracting texture features from a texture image, and then we use improved active contours without edges to segment the texture image. Our method has three advantages. Firstly, it has excellent texture discrimination and low complexity, for the uniform LBP/C operator reduces the number of texture pattern by 77%. Secondly, a multiple object texture image can be segmented, for the proposed method employs the LBP/C histogram to evolve the initial curve during the segmentation process, and is expanded to multiphase segmentation by the hierarchical method which can avoid the problem due to the choice of initial conditions. Finally, the speed of image segmentation is higher, for an additive operator splitting （AOS） scheme is applied to the numerical solution of the proposed method. Experiments on both synthetic and remote sensing images show that the proposed method is more accurate and fast. The segmentation error rate of the synthetic texture image is reduced from 3.8% to 1.7%, and the segmentation speed for the two images is increased by over 40%.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2007年第5期712-715,共4页
Journal of Northwestern Polytechnical University
基金
西北工业大学种子基金(Z200538
Z200737)资助
关键词
纹理图像分割
活动围道
无边缘活动围道
局部二进制模式
texture image segmentation
active contour
active contours without edges
local binary pattern
