An Image Highlights Removal Method Based On Polarization Principle

The existing highlight detection and removal methods are usually effective under strict conditions,such as light sources with fixed directions,specific types of light source and geometry shapes of experimental target.Thus,these methods are not applicable under natural environment,especially when highlight region degrades texture details of object surface.To address this problem,we propose a novel,simple and effective method to detect and remove highlight under natural environment.Based on the principle of polarized light,we first design a polarizer rotation control system.Then,we apply dark channel to automatically detect highlight region.Finally,with the principle of physics,highlight region is removed under natural illumination.In this paper,we use polarizer rotation control system to take three images with different polarized angels.The STOKES parameters are calculated within the highlight regions in the three images.With the STOKES parameters,a polarizer's angel for highlight removal is precisely obtained,and highlight regions can be removed based on the polarizer angel.The highlight removal result is a RAW image,and then we restore the final color image from the RAW image.Experimental results and comparison with other methods show that our method can effectively remove the highlight regions of the target image,and restore image texture details on the highlight regions under natural environment.