Remotely Sensed Image Inpainting With MNLTV Model

Abstract

Image processing is an significant component of modern technologies as it provides the perfection in pictorial information for human interpretation and processing of image data for storage, transmission and representation. In remotely sensed images because of poor atmospheric condition and sensor malfunction (Instrument error such as SLC-OFF failure on may13,2003 the scan line corrector (SLC)of LANDSAT7 Enhanced Thematic Mapper Plus(ETM+)sensor failed permanently causing around 20% of pixel not scanned which become called dead pixels)there is usually great deal of missing information which reduce utilization rate. Remotely sensed images often suffer from strip noise ,random dead pixels. The techniques to recover good image from contaminated one are called image destriping for strips and image inpainting for dead pixels, therefore reconstruction of filling dead pixels and removing uninteresting object is an important issue in remotely sensed images. In past decades ,missing information reconstruction of remote sensing data has become an active research field and large number of algorithms have been developed. This paper presented to solve image destriping , image inpainting and removal of uninteresting object based on multichannel nonlocal total variation. In this algorithm we consider nonlocal method which has superior performance in dealing with textured images.To optimize variation model a Bregmanized-operator-splitting algorithm is employed. Furthermore proposed inpainting algorithm is used for text removal, scratch removal ,pepper and salt noise removal ,object removal etc. The proposed inpainting algorithm was tested on simulated data.