International Journal of Engineering Science Invention Research & Development; Vol. III, Issue VI, December 2016 www.ijesird.com, e-ISSN: 2349-6185 Anupama Sanjay Awati and Meenakshi R. Patil ijesird, Vol. III, Issue VI, December 2016/ 356 DIGITAL IMAGE INPAINTING USING MODIFIED CONVOLUTION METHOD. Anupama Sanjay Awati*,Meenakshi R. Patil** * Dept of E & C, KLS Gogte Institute of Technology, Belgaum, India. **Dept of Electronics and communication, JAGMIT, Jamkhandi, India. Abstract—Digital Image Inpainting is a developing research area in Image Processing. Digital Image Inpainting is a technique in which the damaged part of the image is reconstructed by using the pixel information of the undamaged part. The damaged part of the image can be reconstructed by convolving the image with a kernel function based on nearby pixel information. In this paper a modified convolution method is proposed in which the damaged image is reconstructed by convolving the damaged part with a kernel function by approaching in all four directions. This method gives better results than the method in which the damaged region is reconstructed in only one direction. Keywords—Digital Image Inpainting, Convolution, kernel function, SSIM Structural similarity. I. INTRODUCTION Image Inpainting is the art of restoring the lost or damaged part of an image. This field is very active subclass of image processing. Image Inpainting has numerous applications like photo restoration, error concealment of transmitted images, removing the unwanted text & logos from image, producing stunning visual effects in image. Restoring the unity of the image is the main objective of Image Inpainting. Using Image inpainting we can remove the hole with its background as if the omitted object never existed. Thus Image inpainting is aimed at filling the hole to create a pleasing continuation of information such that the editing cannot be easily identified by a neutral observer. Inpainting does not reconstruct the original image but fills the lost or unwanted objects by a part that has close resemblance to the original image. Image inpainting was traditionally done by artists but was a time consuming process. Image inpainting is a different method then noise removal. Noise carries the information about their underlying data. But on the other hand, in Image Inpainting the lost or the corrupted part carries absolutely no information. M. Bertalmio was the first person to introduce the notion of the Image Inpainting. He introduced a simple algorithm where the missing region was filled using the third order Partial Differential Equation. Presently there are different approaches to digital image inpainting and can be broadly classified into different categories as listed below 1. Partial Differential Equation (PDE) based inpainting 2. Exemplar and search based inpainting 3. Hybrid inpainting The Exemplar based approach is a patch level approach and fills the larger areas properly but fails in restoring the definite shapes. It mainly concentrates on the texture of image and hence the structure of image is damaged. The hybrid model combines the advantages of the PDE and exemplar based approaches and hence both texture and structure are reconstructed properly. Most of Image Inpainting algorithms work as follows. Firstly, the regions to be inpainted are selected by a user manually. Next, known information is propagated from region boundaries inward i.e., the known information surrounding the gap is used to fill the gap. Various approaches have been used by the researchers for digital image inpainting which may be classified into the following broad categories: Partial Differential Equation (PDE) based Inpainting, Texture Synthesis based Inpainting, and Exemplar based Inpainting, Hybrid Inpainting, fast Digital Inpainting and Convolution Based Methods [1]. II LITERATURE SURVEY