基于边缘梯度方向直方图的图像检索算法

在基于内容的图像检索中，不同图像对形状细节的要求不同及形状特征对旋转的敏感
性，影响检索性能。对此，本文提出了一种基于边缘梯度方向直方图的图像检索算法：利用B样条小波的多尺度变换模极大值进行边缘检测，实现不同图像对形状细节不同要求；用边缘梯度方向直方图表示物体的形状，采用循环移位最小值计算直方图的相关性进行图像匹配，使其具有旋转不变性。实验证明，算法对图像的亮度变化、缩放、旋转具有不变性，有较高的检索性能。
关键词：基于内容的图像检索；边缘检测；直方图；小波分析
Abstract：In content_based image retrieval, different details are requied by different images and shape is more sensitive to circumrotation than color and texture ,which decrease image retrieva performance. An algorithm for image retrieval based on edge gradient direction histogram is proposed.Image edge is detected by B-Spline wavelet transform based on maximum model in order for resulting that different details are requied by different images and represented by the histogram of gradient directions at edge pixels.The relativity of histogram which results from itertative rotation minimum matching is used to avoid the sensitivity of rotated image. Experimental results show that the algorithm has the robustness against illumination changes, position translation, scale and rotated image and a superior image-retrieving performance.
Keywords:Content_based image retrieval; Edge Detection; Histogram;Wavelet