利用数据场和欧氏距离的图像边缘提取

Image edge extraction based on data field theory and Euclidean distance

图像边缘是图像分析和识别的基础,图像边缘信息的准确性和完整性对后续图像分析和识别有重要影响。为实现图像边缘有效提取,提出一种利用数据场和图像欧氏距离的图像边缘提取方法。首先,该方法利用数据场理论构建图像数据场,实现图像灰度值特征空间到数据场势值空间的转换。然后,在对图像数据场的势值计算时引入图像欧氏距离,利用图像区域欧氏距离扩大像素差异,抑制微小细节和噪声,得到"背景"和"目标"相对分离的势值图。最后,用改进Canny算法对势值图进行边缘提取。实验表明,用本文方法可以有效提高边缘提取的准确性,减少伪边缘,抑制冗余细节和噪声。

Image edges are the basis of image analysis and recognition, therefore the accuracy and continuity of image edge information can affect subsequent image processing operations such as image analysis and recognition. In order to realize the effective extraction of image edges, we propose an image edge extraction method based on the data field theory and the Euclidean distance. We first construct the image data field based on the data field theory and convert the gray space to the data field potential space. We then employ the Euclidean distance to calculate the potential value of the image data field. The introduction of image Euclidean distance in the calculation of data quality can expand pixel difference and suppress small and trivial details and noise, thus obtaining a potential value graph with separated ＂back- ground＂ and ＂target＂ potential values. Finally we utilize the improved Canny operator to extract the edge. Experimental results show that the proposed method can effectively improve the accuracy of edge extraction, reduce false edges, and suppress redundant details and noise.