Object removal by Kriging Interpolation Technique
Lakshmi Bhanu Bangaru, Vimal Kishore Gupta · 2015
Object removal in an image, is very useful in image processing field. Most often there comes a need to remove some scratches, small objects, unwanted text or logos from images. It can be done manually with a lot of effort and time. So, automatic version of this method is required through which it can be done easily and in lesser time. This can be done mainly in two ways - diffusion based and exemplar based methods. Exemplar based method gets the value of unknown region with the help of known region at the patch level. Diffusion based method can get the values of unknown region with the help of known pixel values. These are pixel level methods. Here one of the diffusion based methods is discussed which is called Kriging Interpolation Technique. Kriging interpolation technique uses the known neighbouring pixel values and their distances to calculate the unknown pixel value. In this method, image is divided into kxk blocks and then kriging algorithm is used on each block to get the required image. User has to provide the mask manually, which means, user has to tell which part of the image is unwanted. This technique gives better results. Result can be analysed with PSNR and MSE values. The most important applications of the image inpaiting are object removal, scratch removal, restoring missing areas, image repairing etc., An image or photograph which is damaged because of aging can be restored by image inpaiting. At present, mask has to be provided by user. In future, it can developed as the mask can be constructed by the algorithm itself.