【摘要】 提出一种基于纹理基元分布统计的纹理分类算法 ,选定一组代表像素变化的基元序列 ,计算每一个基元在纹理图像中的覆盖比例 ,用得到的纹理基元属性分布作为描述参数 ;由于相似纹理其属性也是相似的 ,同类纹理必然有接近的基元分布参数 ,计算参与实验的纹理样本的基元分布的互方差及互相关 ,与代表相似程度的阈值比较判断 ,由获得的共性来锁定同类纹理 ;为使同类纹理具有可参照的标准 ,产生针对每一类纹理的标准类分布。对 Brodatz的 1 1 1纹理不同相似程度的分类结果表明 ,该方法保证了统计结果与视觉判断的一致性 ,可用于纹理的分类及识别。更多还原

【Abstract】 A texture classification algorithm based on statistic result of primitive element distribution is proposed. A sequence of element representing pixel variations is chosen,the proportion occupied by each of the elements is calculated,and the obtained texture element attributions are used to describe a texture image. Since similar textures have a similar attribution,textures of same type have closed element distribution parameters. The corral and variance between each texture distribution with others is calculated compared with a threshold representing the similarity between two textures. And then standard distributions are obtained corresponding with similar texture images in one group. In the experiment,111 Brodatz textures are classified with different similitude degrees. Results show that the algorithm ensures statistical results coincident with the visual judgement.更多还原