基于HVS的自然场景图像特征提取方法

Nature scene image extraction method based on HVS

提出了基于人类视觉系统的自然场景图像颜色和纹理特征提取方法。介绍了两类低层图像特征颜色和纹理特征提取方法,分别提出了以自适应主导色原理来表示颜色特征、采用纹理灰度图的空频域特性表示空间纹理特征;以ACA方法获得自适应主导色、以可控金字塔滤波分解的方法获取纹理特征。同时还介绍了优化颜色构成距离度量和纹理距离度量,最后给出了实验结果和分析。

A nature scene image color-texture feature extraction method based on human visual system （HVS） was proposed. Two types of low-level image features color and texture feature extraction methods were introduced, and adaptive dominant colors principle described color features, spatial and frequency characterristic of texture grayscale component describes texture features, using adaptive clustering algorithm（ACA） to get adaptive dominant colors and steerable pyramid filter decomposition to extract texture features. Meanwhile, optimal colorcomposition distance （OCCD） and texture distance metric approach were also presented. Finally experiment results and analysis were given out.