Bias Removal Blending to Create Panorama

Abstract

Robust feature extraction and image stitching to create a panorama is a challenging task with various applications in computer vision and robotics. In order to describe and get the best conceivable stitching, this paper put forward an approach of panoramic image stitching and blending process which is divided into three steps. SIFT feature extraction, feature matching and image blending to generate a seamless panorama. The stitched images to create a panorama that have significant illumination changes at the stitched line appear to be unnatural. An image blending algorithm is developed in this paper, based on minimizing sharp edge error by a weighted matrix to solve the problem. First, panorama is generated using feature space of SIFT, so that the images are to be stitched should have maximum crucial features, and performs the rough matching process, followed by RANSAC algorithm for fine matches. Finally applied the different image blending techniques between three images and our method successfully minimizes the error to have a uniform illumination over panorama. The results are compared with the two blending processes. Results are demonstrated by visual assessment and quantitatively by calculating normalized edges. Experimental results show that our algorithm is effective and able to make sharp changes disappear at image joins. This method will be very useful for tracking aerial target near launchpad with multiple fixed cameras with same Field of View.