Real Time Exposure Fusion Based on the Polarization Camera

The new sensors with the pixel-level polarizer structure make us able to shoot once and get four images with different light characters at the same time. It is similar to change the exposure four times and creates opportunities to obtain images with higher dynamic range. Based on the character of the novel polarization camera and the image acquisition pipeline of the regular digital camera, we propose a real-time exposure fusion method to reconstruct the high dynamic images. Polarization information is used to construct the base layer as a reference for the calculation of the blending weights in the method. The result will benefit visual navigation as a result of the better input image source. We also compare our method with three exposure fusion methods. Experimental results indicate our comparable performance and lower complexity.