A quick question; why does canny edge detection work better on grayscale images? It is stated everywhere that it works only for grayscale though i tried it for color images and it works on them too, only it detects too many false edges. What is the reason for that, it works on intensity and i believe intensity stays the same whether it is a color image or grayscale?
yes it does if you are trying to write a code for object reognition using contour detection. I believe this has to do with image processing and computer vision? no?
it is a simple question on how the canny algorithm works for the ones who know it. if you dont know the answer for that you can just ignore it and go back to your miserable fucking simple life. Geez every child who learns how to write a simple code in html thinks he's a programmer nowadays.
How are you representing the colour space? Try performing edge detection in HSV instead of RGB (if that's applicable advice. Otherwise, act appropriately.)
Name:
Anonymous2012-07-03 7:50
>>15 ding dong!
You might want to come up with an algorithm for a weighted "average" function() of HSL (possibly with a scale wider than 0-255 or 0-360) and then simply use that as your input for grayscale color recognition.
But good luck getting the weights right - otherwise someone would have done it well already.
Note that it's worth a shot as an exercise.
The weights should probably be calculated dynamically for every image or portions of it... as based on the data available in that part of the picture e.g. if the H channel shows low variance use the more varying channels more agressively and so on.
But you'll probably be putting most emphasis on luminescence/brightness anyway, so straight grayscale will be almost as good as trying to conjur up more info from color.