We present an approach to easily remove the effects of haze from images. It is based on the fact that usually airlight scattered by atmospheric particles is partially polarized.
abstract
We present an approach to easily remove the effects of haze from images. It is based on the fact that usually airlight scattered by atmospheric particles is partially polarized .We analyze the image formation process, taking into account polarization effects of atmospheric scattering. We then invert the process to enable the removal of haze from images.
Project theory overview
Figure 1 – Light coming from the illuminant and scattered by atmospheric particles towards the camera is the airlight
Recently there has been a growing interest in the analysis of images of scenes affected by weather phenomena . There was already some methods to correct the affect by require prior information about the scene , other methods are based on specialized sources and detection hardware . Our approach is based on analyzing images taken through a polarizer. In general , however , polarization filtering alone cannot remove the haze from images. Here , we obtain much more than optics alone can yield by analyzing the polarization filtered image. The principle we use in our project is very simple : The image is composed of two unknown components : the scene Radiance in the absence of haze , and airlight . To recover these two unknowns we need two independent images. We easily obtain these images because usually airlight is partially polarized . The method only requires that the airlight induces some detectable partial polarization . We demonstrate removal of haze effects from a real scene in a situation where pure optical filtering not suffice at all .
Tools
Figure 2 – System-Schematics
The means used in our project workstations are special hardware: Donfisha 3CCD Color camera module, Stepper Motor and Controller . PC with the following software packages : Lab View 6.1 , stepping motor driver and Matlab.
The Results : Dehazed images examples
Example 1 – The contrast of features in the dehazed image is greatly improved relative to both of original images
Example 2
Algorithm has removed the blue color bias, which existed in the raw images. Thus the green fields are clearly visible, while in the raw images that area looks like brown background. The colors of the red distant buildings are also restored. The mountains at the airlight are clearly visible and also a far away buildings on the mountains become visible.
Conclusions and the next step
a) The test bench was made in order to make dehazing measurements and data processing.
b) The algorithm was checked on the images obtained from the test bench. c) The results showed that in general, an image improvement was obtained. d) There was a problem in getting correct sky image. The resulting sky was full of noise and the colors didn’t match the actual sky colors. This happened because of the way that Matlab and Labview interpret numbers over 255.
e) The next step could be measurements of images at different levels of illumination and at different polarizer angles.
Acknowledgment
We are grateful to our project supervisor Dr. Yoav Schechner for his help and guidance throughout this work, and the VISL staff to Yohanan Erez and Yakoby Aharon .
We are also grateful to the Ollendorff Minerva Center Fund for supporting this project.