Pixel-feature-controlling edge detection based on regularization (PEDBOR)

The PEDBOR algorithm, in which a threshold function is used for all images, is presented. The algorithm is iterative. In each iteration, pixels are first grouped into three categories: edge pixels, homogeneous region pixels, and noise pixels based on the magnitude of the gradient and the orientation distribution of the gradient in the neighborhood of each pixel. After grouping, a regularization process is applied to the pixels. Unlike other regularization methods, the PEDBOR algorithm chooses regularization parameters, which control the spatial smoothness of processed images and the fidelity of the processed images to original images, according to the pixel type. The algorithm is suitable for a wide range of images and is robust in a variety of noisy situations.