摘要
将图像中的主纹理提取出来对生成高质量的纹理是很重要的.已有的主纹理提取算法计算复杂、时间开销很大,所知最快的方法处理一幅样本图像也要超过10 min.对此,提出一种简捷的方法,可将速度提高4个数量级,在1 s内处理一幅样本图像,且提取的主纹理及相应的合成结果与已有工作相当.由于主纹理往往在样本中占据大部分的面积,且其色彩特征是反复出现的,为此通过多分辨率的色彩统计直方图,将出现频率高的色彩特征所相应的像素作为主纹理区域.使用HIS颜色模型进行色彩统计,并统计多种分辨率下的样本纹理的色彩分布情况,将大多数情况下可被归于主纹理的部分当作该样本的主纹理.
Existing methods for extracting dominant textures are always complicated and very expensive,e.g.more than 10 minutes are often required to handle an image sample.This paper proposes an easy and fast method,being able to extract dominant textures from an image sample in less than 1 second,and its results are comparable with those of existing methods.The new method is based on the key observation that dominant textures often take up a larger portion in the image sample,and their color features appear frequently.Thus,with multiscale histograms,we may find the color features with high frequencies,and have their corresponding pixels belonged to the dominant textures.We adopt the HIS color model and re-sample the image sample in a few resolutions,and perform the investigation on them respectively,by which the pixels often regarded as belonged to dominant textures will be finally taken to form the dominant textures.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2011年第1期46-53,共8页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金(60773026
60873182
60833007)
关键词
主纹理
直方图
HIS颜色模型
提取
高效
dominant textures
histogram
the HIS color model
extract
high efficiency