Research On Image Completion Algorithm Based On Structure Matching And Propagation

With the progress of the society and the rapid development of computer science and technology, digital image completion technology has been used in many fields. Such as: antiquities’ pictures protection, network transmission, acrobatic show, medicine, military, etc. Based on this background, some digital image completion algorithms have been studied in ths article.When using the classic BSCB(Bertalmio, Sapiro, Caselles, Ballester) image completion algorithm based on PDE(Partial Differential Equation) to complete the damaged images with larger damaged area containing salient structures, the completed images might appear the problems that the completed structures could be fuzzy or the two broken structures belonging to the same could not be connected well. In order to solve the problems of BSCB, images with larger damaged area containing salient structures were completed by simulating the manual repair methods of painter, and three algorithms have been proposed:(1) By using the straight path to connect the well-matched broken structures, the ICSMP-ESPC (Image Completion based on Structure Matching and Propagation using Endpoints Straight Path Connection) algorithm was put forward. The experimental results showed that, the ICSMP-ESPC algorithm solved the existing problems of BSCB, at the same time, it had better image completion effects than some classic references’algorithms.(2) By using the planar smooth path to connect the well-matched broken structures, the ICSMP-SFM (Image Completion based on Structure Matching and Propagation using Smooth-fitting Model) algorithm was presented. This is the improved algorithm of ICSMP-ESPC. The ICSMP-SFM algorithm solved the existing problems of ICSMP-ESPC, that ICSMP-ESPC algorithm is not suitable for completing the damaged images with larger damaged area containing structures with larger curvature.(3) The new constructed UK-flag shaped model was used in the diffusion process of BSCB in the ICSMP-SFM algorithm, the ICSMP-USM (Image Completion based on Structure Matching and Propagation using UK-flag Shaped Model) algorithm was proposed. The Experimental Results expressed that, the ICSMP-USM algorithm has improved the image completion quality of ICSMP-SFM to some extent.