Image edge detection based on local dimension: A complex networks approach

• A novel method is proposed for modeling the image to complex network.
• The proposed method based on local dimension is to address edge detecting.
• It is observed that edge pixels have lower node dimension than non-edge pixels.

A novel approach based on local dimension of complex networks is proposed to address edge detecting issue in image processing. A new method that the image was modeled to complex networks is proposed. Our method for mapping an image into complex network is based on the weighted combination of the Euclidean distance and gray-level similarity indices, which differs from the way was done previously in several works. The local dimension of node characters the local property of pixel. It is observed that edge pixels obviously have lower node dimension than non-edge pixels, which is used in our proposed edge detection method. The proposed method is applied to both synthetic and natural images. The results show the efficiency of our proposed method.