Research On The Cumulation Algorithm Of 3D Reconstruction From Gray-level Image

As a key research direction of computer vision, three dimension passive reconstruction becomes more and more important so as not to be ignored in the various areas of our society and has been widely used in industry, agriculture, country defense, medicine, space technology, etc, especially. Due to its special feature, which can restore 3d shape information just from one gray-level image, recently, shape from shading (SFS) gets great improved and develops rapidly in methodology and application , and becomes a new research focus of 3d passive reconstruction.On the base of comprehensive summarization and analysis of current SFS research work, a 3d reconstruction technology from gray-level image has been proposed in this paper. With the quadrangle composed of four nearest pixels as the least element, a set of nonlinear equations can be obtained from the relationship between the normal and gray of four triangular planes; then, the relative location of four corresponding point of these four pixels can be acquired from the solution of these nonlinear equations with the nonlinear least square method. By this way, the precision of the result of 3d reconstruction is mainly determined by the resolution of this image after all the relative pixels' locations of this object surface are figured out through a cumulation algorithm. Because this new algorithm don't adopt rigor assumptions that don't correspond to the real natural condition and unsuitable assumptions that some prior condition has known as most traditional SFS algorithms do, it can get more precise and universal 3D reconstruction result.Meanwhile, a new rendering algorithm-reverse calculation algorithm-has also been presented by the contrast between the current popular rendering method-GBI IBR PBR in this paper, through which the color of each pixel can be figured out under the lighting model after the coordinates and normal of every corresponding point on the 3d object surface have been calculated from the pixels in the range where the 3d object is projected onto the imaging plane, in order to provide each pixel a real color and get an image with the strongest feeling of reality. By this way, the complexity and error because of the camera calibration can be avoided and a perfect research flat will be acquired.By the initial experiment, on the base of the gray-level image from reverse calculation rendering algorithm, with the cumulation algorithm of 3d reconstruction, a better result has been gotten. Through the contrast between the result and the real data of this 3d object, theerror of element reconstruction is no more than a tenth part of pixel, which is mainly determined by the resolution of this gray-level image.