Research On High Dynamic Range Image Fusion Technique Applied On Space Observation

The dynamic range of imaging devices is often lower than that of natural scenes or human eyes, therefore the images acquired often cannot mirror the real scene, especially the details in the dark and highlight areas. This problem can be addressed by High Dynamic Range Imaging and Multi-exposure Image Fusion. However, these two technologies both are based on the premise that there should not be relative motion between the imaging target and the imaging device.When a space observation experiment is proceeded, the reflectivity of each parts of the target differs, and there exists relative motion between the target and the camera, which makes this problem much more severe. To address this problem, a multi-exposure image fusion system is proposed. In order to extend the dynamic range of the special observation images, image registration module and image correction module are added before the multi-exposure fusion module in the system.In this project, the main work accomplished is as follows:Firstly, a comprehensive and in-depth investigation on image registration and multi-exposure image fusion algorithm has been made at home and abroad. Secondly, an introduction about theories of High Dynamic Range Imaging and storage formats as well as display methods of HDR images have been made briefly. Thirdly, according to the characteristics of space experiments and space images, a multi-exposure image fusion system has been proposed, which can be applied on images with relative motion between the camera and scene. The system is realized by using SURF algorithm to extract features, then proceeding rough-fine matching with algorithms that based on Euclidean distance and RANSAC respectively, correcting and clipping the images through the fitted image transformation models, and extending the dynamic range of images by using multi-exposure fusion algorithm that based on Gaussian and Laplacian pyramids decomposition. Fourthly, the system has been tested by a great deal of experiments as well as subjective and objective assessments, and is proven to be able to achieve the preset goals effectively.The innovation points of this essay are as follows:(1) According to the characteristics of space images, a multi-exposure fusion system that is applied to process the space images is proposed for the first time in order to extend dynamic range.(2) Considering the relative motion between the imaging target and the imaging device in the space experiment, an image registration module based on SURF features and RANSAC algorithm is included in the proposed system to correct distortions in the images.(3) A software platform is designed with C++, which is able to process the images on a graphic interface.