Research On Automatic Generation Algorithm Of ID Photos Image And Trimap For Image Matting

Oriented by the development of ID photo processing software requirements,research on qualified ID photo processing and generation algorithms.In order to process the photos perfectly,explore the image matting algorithm,and study the automatic generation algorithm of the required trimap.Because of the disadvantages of FCM clustering algorithm required by the automatic generation algorithm of trimap,this paper studies how to improve it.The thesis work mainly includes the following points:(1)The TFR algorithm for automatic generation of ID photos is proposed.Aiming at the second-generation resident ID card with strict standards for digital photos,in order to process non-compliant photos into qualified ID card photos,an image processing method for ID card photos based on TFR algorithm is proposed.The TFR algorithm processes the photos that do not meet the standard in three steps:(1)Crop the photo with the head distance cropping of T algorithm,(2)Use the improved Flood Fill of F algorithm to replace the photo background,(3)Finally,use the mean value RGB skin of R algorithm was used to adjust the skin color.Using the ID photo image processing method in this article to process photos,users only need to provide a single-person photo with a simple background,and it can be automatically processed into a standard ID photo,which solves the inconvenience problem caused of the usual need to go to the photo studio to take ID photo photos.(2)A clustering algorithm(KDPC-FCM)is proposed which combines the peak density and FCM of KNN optimization.In order to solve the problems of fuzzy c-means clustering algorithm,which is sensitive to initial clustering center and noise,inaccurate to boundary sample clustering and easy to converge to local minimum,a fusion clustering algorithm combining K-Neighborhood(KNN)optimization density peak(DPC)algorithm and FCM is proposed.The algorithm uses the k-nearest neighbor Information to define the local density of samples,quickly and accurately search the peak density points of samples as the initial cluster center,and improve the shortcomings of FCM clustering algorithm,so as to optimize the effect of FCM clustering algorithm.The experimental results show that the improved algorithm has better noise resistance,clustering effect and faster global convergence speed compared with the traditional FCM algorithm and DSFCM algorithm proposed in other literatures.It proves the feasibility and effectiveness of the new algorithm.(3)The ESD algorithm of automatic generation trimap for image matting is proposed.Image matting is a serious under constraint problem.In order to solve this problem,users need to input in the form of trimap,but manual generation of trimap is a time-consuming and laborious process.In order to solve this problem,this paper proposes an ESD algorithm,which is composed of E algorithm based on DRFI,S algorithm based on lazy snapping and D algorithm based on KDPC-FCM,by combining image saliency,image segmentation and KDPC-FCM clustering algorithms.The experimental results show that,in addition to automatically generating accurate trimap,compared with the previous algorithm,the ground truth generated by the best trimap generated by ESD algorithm contains less artifacts and faster calculation speed,and the proposed method does not rely on image depth information as the previous method.The experimental results show that the proposed method is effective.