This paper proposes a CBIR (content-based image retrieval) method which combines maximum connected region area histogram with visual weights of pixels.At first,the method calculates the color complexity and the visual weight of every pixel.Secondly,the maximum connected regions in the HSV color quantization space are detected.Then the visual weight for the maximum connected region of every color grade is calculated using the visual weights of pixels.Finally,the features of maximum connected region area histogram based on visual weights of pixels are constructed,and suitable image similarity functions are employed.Experimental results show that the retrieval results are more consistent with human visual recognition.%提出一种对最大连通面积直方图进行视觉加权的图像检索方法.该方法首先计算像素点颜色复杂度和视觉权值;然后找出HSV颜色量化空间下各种颜色等级的最大连通区域;再采用像素点的视觉权重,计算连通区域所对应的视觉权值;最后构造出视觉加权的最大连通面积直方图特征,并利用适于该特征的相似性度量函数实现检索.实验表明检索结果更加符合人的视觉感知.
展开▼