基于图像的动物皮革纹理提取

Texture extraction based on leather image

动物皮革表层的纹理具有其自身特色,对纹理的提取不仅可用于皮革类别的鉴定,也有助于对自然纹理的再设计和应用。直接采集的皮革图像,纹理模糊、粗细不匀且含有色泽差异。在对动物皮革纹理总结的基础上,运用微分算子、Canny边缘检测算子以及分水岭算法对皮革图像进行了纹理提取,并对不同方法进行了比较分析。试验结果发现:不同算法效果存在着显著差异,其中Canny算子能够有效地提取出主干纹理,分水岭算法则会将纹理细节充分表达出来。Canny算子及分水岭算法的引入,能够有效地提取出皮革图像的纹理信息,为人造皮革的纹理仿制提供了自然纹理参照。

Animal leather owns special texture feature which can be helpful for the leather identification, the re - design and application of the natural texture. As blur texture on the surface of leather, captured image shows thick or thin texture and color difference. After summarizing the texture characteristic of leather, differential operator such as canny edge detection operator , wa- tershed algorithm were used to extract the texture of the leather image, and were compared with each other. The results show that there are significant differences between different algorithms. Canny operator can effectively extract the main texture, while the wa- tershed algorithm can fully express the texture details. The use of this method can effectively extract the texture information of the leather image, and provide the reference for the art design with natural texture.