Exemplar Based Image Inpainting Algorithm Using Structure Tensor

International Journal of Research and Scientific Innovation (IJRSI) | Volume V, Issue VI, June 2018 | ISSN 2321–2705

Exemplar Based Image Inpainting Algorithm Using Structure Tensor

Saraswati¹, Dr. Prashant P.Patavardhan²

^{1, 2}Department of Electronics and Communication Engineering, KLS Gogte Institute of Technology, Belagavi, Karnataka, India

Abstract: Image inpainting is a technique to fill the missing area in the image, to recover the damaged area in the image and to remove unwanted area in the image in an undetectable way. After applying inpainting method to an image the output image should look like original image and The viewer should feel that the image is never altered before. In this paper the proposed algorithm is based on image structure tensor to tackle the problems of criminisi’s algorithm. Image gradients with image structure have been employed to help image structure detection. Performance analysis of the structure tensor is shown in result section comparing with crimnisi’s algorithm by using PSNR and SSIM and also the runtime of the both algorithms.

Keywords: image inpainting, criminisi’s algorithm, structure tensor, image gradient, structure analysis.

I. INTRODUCTION

In real world, image Inpainting is very hottest topic in the image processing to fill the missing area in the image, to recover the damaged area in the image and to remove unwanted area in the image. the viewer cannot find the image is altered by Inpainting technique. The Inpainting algorithm is based on known area of the image image look like there is no visual differences between original image and Inpainted image. Inpainting algorithm has several applications such as removing unwanted objects from image, image compression, restoration of an image, filling missed region. techniques of image Inpainting are texture synthesis based image Inpainting[13] algorithm is the earliest techniques of image Inpainting in image processing. To fill the missing areas texture synthesis based algorithms utilize same neighborhoods of the damaged pixels in the image. The earlier Inpainting techniques make use of texture synthesis methods to fill the damaged area by sampling and copying pixels from the neighboringpixel. PDE based image Inpainting[13] the main goal of partial differential equation (PDE) algorithm is to inpaint the image partial differential equation algorithms outputs are good if inpainting area is small for large inpainting area it will take more time and output will look like blurry image.

Read More