An Algorithm Of Texture Mapping For3D Models Based On Boundary Adjustment And Image Blending

Recently, the3D reconstruction, especially the texured modelreconstruction is of increase importance and more widely applied indifferent fields. So in this paper, we focus on the method of texturemapping in the3D reconstruction, and propose an algorithm of texturemapping based on boundary adjustment and image blending. The newalgorithm aims at solving two kinds of problems in the process of texturemapping.First kind of problem is shown as “double texture” or “texturemismatch” in the textured model, and results from the match error betweenthe3D model and the texture image, and also the estimation error of thecamera pose. Considering the problem, we define the texture cost functionand edge cost function to build a optimization problem. And anapproximate solution is acquired by some image processing method. Thenwe find the optimal position for the boundary of texture mapping. Usingboundary adjustment strategy, this kind of problem is solved.Another kind of problem is the seams in the textured model caused bythe illuminant difference and color difference among the texture images.Because the texture images are taken in common environment when wescan the object, it is difficult to control the illuminant condition fordifferent views. The texture blending strategy based on tone mapping andsmoothing is proposed to generate the texture map with less texturediscontinuity. Tone mapping is first applied to unify the tone of differenttexture images, and smoothing nearly eliminates the seams. Experimental results demonstrate the effectiveness of the proposedalgorithm. It can solve the mentioned problems and improve the quality oftexture mapping.