An effective edge extraction method using improved local binary pattern for blurry digital radiography images

Aiming at an efficient edge extraction method for blurry digital radiography (DR) image of large size, local binary pattern (LBP) is improved by embedding an H function and a counting scheme, so-called H-LBP. With the aid of similarity distance and the H function, H-LBP can extract surrounding information discriminatively. With the help of counting scheme that records the number of pixels whose values are equal to or greater than that of the centric one, H-LBP can reduce noises and blurring. Experiments on blurry DR images and other images show that, compared with the Canny method, a fusion method called FL-fusion in this paper, and LBP operator, H-LBP extracts edges containing much more details.

► We improve local binary pattern to extract edges in blurry DR image.
► We embed an H function and similarity distance into LBP.
► We add a counting scheme to reduce noises and artifacts.
► Our method can extract edges of more details in nearly 4 s.
► Improved local binary pattern is also effective for other real images.