Hello,
I'm getting the difference between two images using the absdiff method and aplying a threshold. However, due to light variations, I have to change the threshold value every place I test my system.
Is there an alternative way to get the difference between two images that is light independent?
Any tip will be very helpful,
Thanks
For python3.5 and OpenCV 3.3.1 on raspberry pi3 using linux. I used createCLAHE and or cv2.equalizeHist or even create function of gamma.
Have you tried normalizing the images by the standard deviation?
Convert them both to the CV_32F type, and calculate the mean and standard deviation of each image. Subtract the mean from each then divide by the stddev.
This is a simple way. If you want to search for more complicated ways, the key terms to use is "illumination invariance". There are a huge number of methods, so good luck.
Asked: 2017-10-29 06:43:42 -0600
Seen: 803 times
Last updated: Oct 30 '17
