3D Object View Construction from 2D images

Abstract

We have constructed 3D images from a pair of 2d Images using a stereo technique as Computer vision and augmented reality industry focus on reconstructing 3D view of an object to improvise the visual effect. There are various fields in which there is a need for 3D images like street views, film industry, game industry, etc. Our research focuses on creating 3D images from pair of 2D images through calibration and reconstruction in less computation time which we have achieved very much as compared with previous work our method takes very much little running time. 3D constructed images results are also quite impressive but some speckles are there still. we have used the chessboard pattern images dataset for camera calibration and Middlebury stereo data for reconstruction through the use of feature match method (SGBM) which matches the correspondence points between pair of images to sub-pixel accuracy. The disparity map technique increases the efficiency and reliability of the reconstruction method. The final 3D constructed images can be visualized in mesh lab software. We have done comparative analysis which proves the improvement in computation time on the same dataset our methods perform very efficiently which is the utmost outcome of our research