Abstract: The local binary pattern （LBP） can extract fuzzy radiographic image texture, but is unable to describe the degree of grayscale difference and distinguish redundant micro changes. To settle this problem, a new kind of rapid texture extraction algorithm C-LBP was put forward. First of all, the complex shock filter was used to preprocess our image. And then the imaginary data were extracted as the next step input. The LBP was improved by a smooth compare function.In order to describe the grayscale informations distinctively, the similarity distances between surrounding points and centric point were considered. Besides, a counting scheme was utilized to eliminate redundant micro-changes. The experimental results show that the C-LBP has advantages in both processing time and detecting vision. It can keep the sensitivity to the large gray level and enhance the description capacity of the grayscale difference. The C-LBP is feasible for image enhancement, edge extraction and defect classification.