摘要： 在图像检索领域中，为更好地利用Julesz基元概念来描述图像内容，提出一种新的图像特征描述方法——局部结构直方图。将彩色图像从RGB空间转换到HSV颜色空间，分别提取边缘方向和颜色信息。定义5种基元类型进行局部结构检测，并利用直方图描述图像特征。实验结果表明，与LBP直方图和Gabor滤波器方法相比，该方法的检索性能较优。

关键词: 图像检索, HSV颜色空间, 颜色量化, 边缘方向, 局部结构检测, 局部结构直方图

Abstract: In order to make better use of Julesz’s texton concept to describe the image content for image retrieval, this paper presents a new method, namely Local Structure Histogram(LSH), to describe image features. It converts color image from RGB color space to HSV color space, and extracts colors information and edge orientation in HSV color space. It defines five textons types for local structures detection, and uses histogram to represent image features. Experimental results show that the performance of the method is better than that of local binary pattern histogram and Gabor filter method.

Key words: image retrieval, HSV color space, color quantization, edge orientation, local structure detection, Local Structure Histogram(LSH)

中图分类号: 

• TP391