摘要
在传统纹理识别方法中,纹理识别是对图像分块进行的,因此得到的纹理边界不光滑,同时呈现明显的分块痕迹。一种充分利用图像邻域信息的"相关度计算"方法,通过计算包含象素点的所有图像分块被识别的概率,来判断象素点是否被识别。结果表明,"相关度计算"方法可以得到相对光滑的纹理边界。
Texture recognition is implemented by separating the image into blocks in traditional methods. So the edge of the texture is not smooth enough and appears an obvious track of block-separating. A method named correlativity calculation was introduced. Making full use of neighborhood information of the image, this method determined whether or not a pixel would be recognized through calculating the recognition rate of every image block which contained the same pixel, and thus resulted in a relatively smooth edge.
出处
《计算机应用》
CSCD
北大核心
2004年第12期25-27,共3页
journal of Computer Applications
关键词
纹理识别
纹理分割
边界
邻域信息
texture recognition
texture segmentation
edge
neighborhood information
