Regional growth inpainting strategy for depth image

Abstract

The depth images acquired from depth sensors have inherent problems, such as missing depth values and noisy boundaries. In this paper, an inpainting strategy for depth image based on regional growth criterion is proposed. In terms of image inpainting sequence, based on Criminisi priority method, a new calculating method of confidence item is defined and the improved priority is used to weight the confidence items and data items. Compared with the inpainting results of Joint bilateral filter (JBF), JBFC (JBF based on Criminisi priority), and JBFW (JBF based on weighted pixel priority), the inpainting sequence and texture extension determined by weighted pixel priority are better. In terms of image inpainting field, Under the guidance of the weighted pixel priority, the adaptive neighborhood of the pixel to be inpainted is defined by region growth criterion which have higher similarity with the pixel to be inpainted than by traditional neighborhood. Experimental results show that the neighborhood constructed by region growth criterion is accurate and effective. Generally, the regional growth inpainting strategy can obtain higher inpainting accuracy while keeping the boundary information