Abstract

The in-vehicle black box camera (dashboard camera)has become a popular device in many countries for securitymonitoring and event capturing. The readability of videocontent is the most critical matter, however, the content isoften degraded due to the windscreen reflection of objects inside. In this paper, we propose a novel method to removethe reflection on the windscreen from in-vehicle black boxvideos. The method exploits the spatio-temporal coherenceof reflection, which states that a vehicle is moving forwardwhile the reflection of the internal objects remains static.The average image prior is proposed by imposing a heavy-tail distribution with a higher peak to remove the reflection.The two-layered scene composed of reflection and back-ground layers is the basis of the separation model. A non-convex cost function is developed based on this property andoptimized in a fast way in a half quadratic form. Experi-mental results demonstrate that the proposed approach suc-cessfully separates the reflection layer in several real blackbox videos.