Seamless Color Mapping For 3d Reconstruction With Consumer-Grade Scanning Devices

Virtual Reality provides an immersive and intuitive shopping experience for customers. This raises challenging problems of reconstructing real-life products realistically in a cheap way. We present a seamless texturing method for 3D reconstructed objects with inexpensive consumer-grade scanning devices. To this end, we develop a two-step global optimization method to seamlessly texture reconstructed models with color images. We first perform a seam generation optimization based on Markov random field to generate more reasonable seams located at low-frequency color areas. Then, we employ a seam correction optimization that uses local color information around seams to correct the misalignments of images used for texturing. In contrast to previous approaches, the proposed method is more computationally efficient in generating seamless texture maps. Experimental results show that our method can efficiently deliver a seamless and high-quality texture maps even for noisy data.