Laser scanning data inpainting method based on tensor voting technique

Laser scanning works by rapidly firing laser beams downwards to the ground to measure the time which the light takes to reflect from objects located on the ground, and then returning to the laser scanner. It is an active method to directly obtain high-precision three dimensional data and has become an established technique for acquiring geometric information. Due to several reasons of the imagery procedure, the acquired image data are often incomplete, which causes the difficulty for the tasks such as feature extracting, image segmentation and object recognition. Considering that tensor voting techniques can provide the geometric information for an image, a laser scanning inpaniting method based on tensor voting technique is described. First, an inpainting prior level is established, second, the damaging point is inpainted based on the percentage of valid and invalid pixels by an iterative scheme for the highest prior level. Because the window size involved in the inpainting procedure is adjusted automatically, so this method is banausic in practice. The experimental results show that the proposed method has favorable inpainting performance compared with the common inpainting method.