基于颜色自相关图的区域定位图像检索

Image retrieval of region location based on color auto-correlogram

颜色自相关图表示了颜色的空间相关性,在图像检索方法中既有效且计算量小,但是该特征在检索前景较为清晰或背景具有较大面积单色的图像时误检率较高.针对此问题,本文提出了一种基于颜色自相关图的区域定位图像检索算法.该算法使用HSV颜色空间自相关图作为图像的底层特征,通过有效区域定位和二值位图来获得局部特征.最后,综合两种特征进行相似度量.实验结果证明,本文方法具有较高的检索精度,克服了颜色自相关图的片面性,显示了组合特征的有效性.

Color Auto-Correlogram distills the spatial correlation of colors, and is both effective and inexpensive for content-based image retrieval. But, this feature has a high rate of false detection when images have a clear outlook or a large area monochrome. For this problem, a new region location based Image Retrieval algorithm based on color co-occurrence histogram was proposed. The algorithm uses HSV color auto-correlogram as basic characteristics, using effective region location and Binary bitmap to obtain the local features. Finally, the features were integrated and retrieved by the best similar matching function. Experiments indicate that the method results from the combination of basic and local features is superior than using single feature.