IMAGE INPAINTING USING NONLOCAL TEXTURE MATCHING AND NONLINEAR FILTERING

Abstract

Nonlocal texture similarity and local intensity smoothness are both essential for solving most image inpainting problems. In this paper, we propose a novel image inpainting algorithm that is capable of reproducing the underlying textural details using a nonlocal texture measure and also smoothing pixel intensity seamlessly in order to achieve natural-looking inpainted images. For matching texture, we propose a Gaussian-weighted nonlocal texture similarity measure to obtain multiple candidate patches for each target patch. To compute the pixel intensity, we apply the-trimmed mean filter to the candidate patches to inpaint the target patch pixel by pixel. The proposed algorithm is compared with four current image inpainting algorithms under different scenarios including object removal, texture synthesis, and error concealment. Experimental results show that the proposed algorithm outperforms the existing algorithms when inpainting large missing regions in images with texture and geometric structures.

Existing System

Partial differential equations (PDE)

Proposed System

In this paper, we propose a new image inpainting algorithm that uses a novel nonlocal texture similarity measure to select several candidate patches for a given target patch, which are then fused together using the trimmed mean filter to obtain the inpainted results. The objective of the proposed algorithm is to recover a large missing region surrounded by multiple textural and structural regions. The proposed algorithm is capable of reproducing the underlying textures and structures of the large missing regions within an image simultaneously and achieving natural-looking inpainted results. There are several contributions of the proposed algorithm. First, this work presents an exemplar-based inpainting method using a new nonlocal texture similarity (NLTS) measure to match the target and candidate patches in the image under the assumption that there are many local repetitive textures present within the image. Next, when choosing the next target patch to be inpainted, the proposed algorithm selects a patch centered on the contour surrounding the outer border of the missing region, unlike other exemplar-based inpainting algorithms. Also, the proposed algorithm chooses multiple good candidate patches and combines them using the trimmed mean filter to inpaint the target patch in comparison to other exemplar-based inpainting algorithms which choose only a single candidate patch. Furthermore, in our proposed iterative framework, the information from the already inpainted patches can be used towards inpainting other target patches as the inpainted pixels are added to the source region, whereas other inpainting algorithms use only the original pixel data. The proposed algorithm is compared to several existing image inpainting algorithms for different applications including object removal, texture synthesis, and error concealment. Both quantitative and qualitative results show that the proposed method outperforms the other inpainting methods. 

Architecture

BLOCK DIAGRAM