计算机工程与应用 ›› 2016, Vol. 52 ›› Issue (10): 181-186.

• 图形图像处理 • 上一篇    下一篇

基于颜色和梯度方向共生直方图的图像检索

谢  莉1,成  运1,曾接贤2,余  胜1   

  1. 1.湖南人文科技学院 信息科学与工程系,湖南 娄底 417000
    2.南昌航空大学 软件学院,南昌 330063
  • 出版日期:2016-05-15 发布日期:2016-05-16

Image retrieval based on color and motif gradient direction co-occurrence histogram

XIE Li1, CHENG Yun1, ZENG Jiexian2, YU Sheng1   

  1. 1.Department of Information Science and Engineering, Hunan University of Humanities, Science and Technology, Loudi, Hunan 417000, China
    2.School of Software, Nanchang Hangkong University, Nanchang 330063, China
  • Online:2016-05-15 Published:2016-05-16

摘要: 为有效融合图像的形状、颜色等特征,提出一种基元梯度方向共生直方图和颜色直方图的彩色图像检索算法。首先将图像划分为4×4互不重叠的基元,并根据基元的梯度平均幅值把基元分为视觉均衡基元和视觉非均衡基元;接着计算视觉均衡基元的平均颜色值,并将其在HSV空间中量化到72色颜色空间,构建视觉均衡基元颜色直方图作为彩色图像的局部颜色描述子;然后将图像中各像素点颜色值在HSV空间量化到72色颜色空间,获得图像的全局颜色直方图;最后在视觉非均衡基元中构建梯度方向共生直方图描述图像的形状特征。融合局部颜色描述子、全局颜色直方图和形状特征构成彩色图像检索特征矢量。实验结果表明,所提算法能够准确描述彩色图像的颜色和形状特征,具有很好的旋转不变性和尺度不变性。相似性度量非常有效,查全率和查准率均有较大提高。

关键词: 图像检索, 颜色直方图, 基元梯度方向共生直方图, 视觉均衡基元, 视觉非均衡基元

Abstract: In order to fuse shape and color features, this paper presents a novel color image retrieval algorithm based on motif gradient direction co-occurrence histogram and color histogram. Firstly, divide the image into 4×4 non-overlapping motif. Based on the average amplitude of the gradient, each motif is classified into visual uniform motif and visual non-uniform motif. Secondly, calculate average color during each visual uniform motif, and quantify it into 72 levels in HSV space. Construct the visual uniform motif color histogram as the local color image descriptors. Then, obtain the global color histogram by uniformly quantize the HSV color image into 72 colors. Lastly, obtain the motif gradient direction co-occurrence histogram to depict the shape feature of image. A color image retrieval feature vector is fused by the local color image descriptor, the global color histogram and the shape feature. Experimental results indicate that the algorithm can describe the color and shape feature, and it has invariance in rotation and size. Its similarity measure is very effective, and it has higher precision and recall rate.

Key words: image retrieval, color histogram, motif gradient direction co-occurrence histogram, visual uniform motif, visual non-uniform motif