基于边缘方向直方图相关性匹配的图像检索

Image retrieval based on edge direction histogram correlation matching

针对基于边缘方向自相关图的图像检索算法的优缺点,提出了一种基于边缘方向直方图相关性匹配的图像检索算法。使用自适应中值滤波器滤除图像中的椒盐噪声。利用Sobel算子提取图像边缘,通过计算边缘梯度幅值、角度统计后得边缘方向直方图,并对直方图进行等级化排列构成特征向量。最后使用斯皮尔曼等级相关计算图像特征向量间的相关系数作为衡量图像间相似性的指标。实验结果表明:该算法的平均查准率、查全率较基于边缘方向自相关图算法分别提升10.5%,9.7%,平均检索耗时减少了7.5%。实验验证了算法的有效性,可将算法应用到中大规模图像检索系统中以提升检索效果,提高系统速度。

With regard to the advantages and disadvantages of image retrieval algorithm based on edge orientation autocorrelogram, a kind of image retrieval algorithm based on edge direction histogram correlation matching was proposed. Firstly, the salt and pepper noise in image was filtered by using an adaptive median filter, and then Sobel operator was used to extract image edge. After the edge orientation histogram was got through calculating the edge gradient amplitude and angle, the feature vector was constituted. Lastly, Spearman rank correlation coefficient was used to calculate the correlation coefficient between the feature vectors of images, as a measure of image similarity. Compared with the algorithm based on edge orientation autocorrelogram, the average precision and the recall rate of the new image retrieval algorithm increased by 10.5% and 9.7%. And the retrieval time consumption was also reduced by 7.5%. The experimental results verify the effectiveness of the proposed algorithm. The algorithm could be applied in medium to large image retrieval system to improve retrieval effect and raise the system speed.