摘要
针对目前纹理图像分割中存在的精度不高问题,提出一种基于局部二元模式方差和四叉树分割的纹理分割方法。首先,利用局部二元模式和方差的比值来计算纹理特征;其次,使用四叉树分割对纹理特征进行分割和合并并在此过程中使用双阈值进行控制;最后,使用圆缓冲区对分割后的边界进行边界光滑。经过对纹理镶嵌图的分割实验,本方法能够较好地分割纹理图像,平均精度在89.6%以上。
Considering the problem of low texture segmentation accuracy by existing methods,a new texture segmentation method based on local binary pattern and quadtree segmentation is proposed.First of all,the texture characteristic is quantified by the ratio of uniform pattern and variance.Then,split and merge step is conducted by quadtree segmentation for the texture characteristics,which by dual threshold control in the process of segmentation.Finally,the use of circular buffer to the segmented boundary smooth boundary is carried on.Through the texture segmentation experiments,it is found that the proposed method can segment the mosic texture image well,and the average accuracy is above 89.6%.
出处
《现代测绘》
2015年第5期12-14,共3页
Modern Surveying and Mapping
关键词
局部二元模式
方差
四叉树分割
纹理分割
local binary pattern
variance
quadtree segmentation
texture segmentation
