A gradient-domain image enhancement method for traffic signs in nighttime surveillance

Abstract

Nighttime video surveillance may suffer from nonuniform illumination from artificial light sources. Meanwhile, these light sources which often result in so noticeable amounts of glow that the objects nearby the light sources cannot be seen at all. To solve the above problem, our paper proposed a novel gradient-domain image enhancement method for traffic signs in nighttime surveillance. We found that both of the glow-removal and image fusion can be represented as a kind of gradient adjustment. Therefore, in our method, remove the glow effects via image gradient decomposition at first. Then, complementary details of inter frames can be fused together in the gradient domain. Finally, the enhanced image will be obtained via a Poisson solver. We have done many experiments and the results show that the proposed method can effectively enhance the traffic signs in nighttime traffic videos.