Adaptive pre processing and image stabilization on embedded platform

The paper presents a unified framework that aims at stabilization of real time images captured from a photosensitive sensor mounted on any mobile platform. Typically the real time images obtained from a mobile platform are prone to suffer from unwanted stray noise, presence of jitter, camera distortions which lead to suppression of finer details and thus make extraction of required information about target a difficult task. The proposed paper presents novel and robust preprocessing stabilization operations that can be implemented on the target image before using it for further processing. Thresholding is one of the most important and challenging operations that is used to extract the crucial finer details of an image and use for application specific purpose. An adaptive thresholding algorithm that segments the target object for both far field and near-field image conditions is proposed. The paper presents the solution to problems caused by non-ideal nature of the lens by carrying out camera calibration in order to achieve un-distortion. The process of camera calibration is carried out using Intel´s Image Processing library OpenCV. The contribution of the proposed work is thus threefold (1) Efficient preprocessing to stabilize image and reduce noise and jitter (2) Improve the robustness by extensive validation on multiple real time images (3) Illustrate the dynamic nature of the proposed framework by customizing it for diverse application scenarios.