The Research Of Generating Portrait Relief Model Based On Single Image

Portrait relief modeling technology is an important 3D model reconstruction technology.At present,relief sculpture is a kind of carving art,which attaches carving elements to the same material background and becomes a highly collectible craft artwork.The portrait relief,with its rich visual aesthetics and unique carving form as a typical representative form of relief art,has unique artistic appeal and commemorative significance.Therefore,in order to meet people’s requirements for customized portrait relief with various styles,the combination of computer graphics and relief art has become a hot topic in the field of computer graphics.The generation of digital reliefs is assisted by 3D modeling technology,which avoids the complicated process and maintenance costs in the traditional relief production process.Aiming at the shortcomings of the actual three-dimensional model data being difficult to obtain and difficult to scan,this paper studies a portrait bas-relief reconstruction technology based on monocular images,and on this basis,designs a model compression method that retains the details of the characters to achieve the Fine modeling of portraits.The main research contents of the paper are as follows:(1)Background reconstruction method based on residual homogeneity and Lambertian reflection modelThe original input image has problems such as blurred details and poor quality,which affect the accuracy of subsequent model construction.Therefore,this paper studies a detail enhancement method based on the homogeneity of image residuals,which describes the detail layer of the image by learning the homogeneous structure of the sampled image and the original image to satisfy the similarity constraint,so as to effectively ensure the quality of the image and the continuity of texture features.At the same time,in order to ensure the continuity and richness of the texture structure of the image foreground face region,after the image enhancement operation,this paper divides the preprocessed image into the face foreground region and the head background region.Aiming at the background area of the image,the Lambertian reflection model is used to estimate the height of the image pixels,thereby restoring the shape of the model.Lay the foundation for the generation of the subsequent portrait relief background area.(2)Face foreground modeling method based on feature point mapping and model compressionFor the foreground region of the segmented image,a model generation method based on feature point mapping is studied.This method maps the pixel features of the image to the feature vertices of the 3D model,constructs a linear combination equation of texture and features by extracting 68 key points of the image,and aligns the corresponding grid vertices of the 3D model to generate a foreground face model.Aiming at the generated foreground model,a non-linear compression method is designed to effectively reduce the loss of model texture details in the compression process by constructing an energy equation containing detail constraints and height constraints.The experimental results show that the method reasonably reconstructs different face models,and has certain universality and scalability.(3)Portrait relief generation method based on model fusion and surface optimizationAfter reconstructing the foreground face model and the background model,the face source model needs to be attached to the surface of the background target model and fused,through the grid parametric mapping of the source model and the target model and fitting a nonlinear height function based on sampling points Method to build the final portrait model.The experiment verifies a variety of portrait images,and obtains portrait relief models that retain their shape and details.The average significance of the algorithm generation model in this paper is 0.685,which is an average improvement of 0.334 compared to other algorithms.Therefore,the portrait relief result generated by the algorithm in this paper can highlight the saliency of the visual area,possess a certain sense of reality,and is suitable for simple portrait images.This research proposes a method to reconstruct a portrait relief model by dividing the foreground and background regions of the image,which is achieved by using only a single image as the input to reconstruct the relief model.Compared with other algorithms,this method can better preserve the portrait texture of the model and highlight the visually significant features of the face region,while this research method can provide the corresponding model and theoretical support for the industrial production of portrait relief products.