Research Of Texture Mapping Based Deblurring And Border Adaptation HSV Local Adjustment

Heritage Digitizing is a very important research direction when computer vision is applied to heritage of culture,among them,there-dimensional reconstruction and texture mapping are widely-applied technology.Practical application scene is background for this thesis.The response to there exists improvement space in current texture mapping,opening several good approaches.And this provides heritage with stronger technical support.Firstly,this thesis gives an introduction on project background and existing technique,development of three-dimensional reconstruction,presentation model for texture,display model for texture and texture mapping.Then states depth image estimation from single image and principle of image deblur.And proposes a new method to deblur single image by estimating depth image from reconstructed mesh,original images and correspond camera parameter matrix accompany with SML operator,in detail,partition image to small block and acquire each Gaussian deblur kernel,deblur each block with its own kernel.Later,this thesis discuss principle of Poisson Equation,RGB color model and HS V color model.Then presents a way to adapt border strip width according to texture patch's characters,and solve Poisson Equation under channel V of HSV color model with border strip as boundary conditions.Finally introduces the final system combined two methods above.Estimation of depth map from constructed model is more accurate than normal ways,and deblurred image brings better result to texture mapping.And the modified local color adjustment works in texture mapping.System like this can make the progress of heritage digitizing much simpler and less painful.