基于改进的颜色和形状特征融合的图像检索

Image Retrieval Based on Improved Color and Shape Feature Fusion

目前图像检索中关于图像颜色的研究大多是独立的研究,并且颜色特征检索的准确率不高,检索时间比较长.针对上述问题,提出一种新的颜色特征量化算法,该算法首先划分颜色的主色调得到8维的颜色直方图,然后和基于HSV空间的前三个低阶颜色矩得到9维颜色直方图进行颜色特征融合,最终得到17维的颜色直方图,用得到的17维颜色直方图方法检索图像.实验证明该检索方法不仅提高了检索查准率和查全率,而且进一步缩短了检索时间.最后这种方法与基于修正的Hu不变矩的形状特征融合进一步提高了检索效率.

At present, most of the researches on image color in image retrieval are independently researched, and the accuracy of color feature retrieval is not high, and the retrieval time is long. In order to solve the above problem, a new color feature quantization algorithm is proposed. Firstly, the main color was divided by using the algorithm and the 8 dimensional color histogram was obtained. Then, the 9 dimensional color histogram of the first three low order color moments based on HSV space were fused to obtain color histogram of 17 dimensions, and the resulting image was retrieved using the 17 dimensional color histogram method. Experiments show that the retrieval method not only improves the retrieval accuracy and recall rate, but also reduces the retrieval time further. Finally, this method is fused with the shape feature of the modified Hu invariant moment, and the retrieval efficiency is further improved.