Space object image enhancement based on intrinsic image decomposition

Abstract

Space object image enhancement is an important problem in computer vision, which is also a fundamental issue for space observation. Instead of single image enhancement methods, we introduce a method based on intrinsic image de-composition by using an image sequence to solve the problem of space object image enhancement. We first capture several space object image sequences via ground simulation and 3ds Max digital simulation. The sequences are under different lighting conditions including changes of the camera exposure time and the lighting directions. Then the intrinsic image decomposition is applied to the sequences and finally we can obtain the en- hanced images which contain more information. Experimental results on our dataset show that the enhanced images are better than the input sequences both in qualitative visual effects and quantitative criteria, e.g. image entropy.