Generating the Super-resolution Image for the Video from the In-vehicle Data Recorder

Abstract

The goal of this research is to generate high-resolution images in ROI by the low-resolution video from the In-Vehicle Device Recorder (IVDR). First, we decide the search region and trajectory of the feature through the construction of the camera model and analysis the imaging geometry and characteristics in the moving camera. Next, we consider the image perturbation and blurring caused by camera movement and try to reduce the impact on image quality. Using the projection geometry can track the trajectory of the feature points movement in the video. An accurately point position estimation in different resolutions requirement is made by the phenomenon that the image will be enlarged over time. We use the patching method to create the high-resolution image in ROI. The proposed rebuilding method for super-resolution imaging is based on motion characteristics of spatial domain features. This method can avoid the problem that generates ring noise using the traditional frequency method. It also has the advantage of simple computing.