0

I want to do motion detection and block matching based on features on the road.

The problem is the lighting effect at night that comes from reverse light mirror shape. The mirror cause the lighting to create lines and not-well distributed illumination. Thus, my algorithm cannot detect the feature motion in next frame.

I have tried CLAHE, basic image pre-processing and averaging illumination. But nothing really works.

I would appreciate any help in removing the effects of the uneven illumination.

enter image description here

Cris Luengo
  • 55,762
  • 10
  • 62
  • 120
Nurul Hidayah
  • 3
  • 3
  • The image is attached with the link above – Nurul Hidayah Aug 16 '18 at 06:15
  • How did you average the illumination? Temporal averaging seems to be the right approach here. Please share some of your results so we don't propose things you have already tried. – Cris Luengo Aug 16 '18 at 09:10
  • For averaging the illumination, I follow this thread : https://stackoverflow.com/questions/9292726/how-to-correct-uneven-illumination-in-images-using-matlab on how to correct uneven illumination.. But, the effect (white line and black area) is still there and features on road still hardly detectable – Nurul Hidayah Aug 19 '18 at 10:51
  • Denoise might remove the noise, but causing the image to blur.. Thus, hardly to detect any feature.. I tried to sharpen, increase the image contrast, but the effect still there.. – Nurul Hidayah Aug 19 '18 at 10:57
  • The main problem is, the lighting from reverse light that cause a very bright white lines and black triangle-shape that appear on every frames.. This possibly cause the algorithm detect that there's no motion even the car is moving – Nurul Hidayah Aug 19 '18 at 10:59
  • Will it be possible to detect features from the image? Any idea? – Nurul Hidayah Aug 19 '18 at 10:59

0 Answers0