opencv - Canny Edge vs Thresholding for contour estimation in Open CV -
i using open cv image processing application involves contour estimation in images. know whether thresholding image (like how have done here) or using canny edge algorithm (here) yields better result. involve algorithmic analysis or missing something?
canny edge detection obviously. whole bunch of things ensure strong edges come out of result. thresholding takes @ intensities , sees whether or not each value smaller or larger , "edge" points respectively. however, depending on complexity of scene, thresholding , edge detection yield same thing. example, if had clean image multiple crisp objects have clear intensity difference between foreground , background, either edge detection or thresholding work. if had more complex image contrast different in different areas, or if had multiple objects different intensities, thresholding not give results because inevitably including in pixels don't belong proper objects. why edge detection better, it's local operator, , thresholding global. thresholding applies set principle every single pixel in image. edge detection decomposes image patches , figures out whether happening in each of patches.
if want take out of this, difference between them both thresholding more used object extraction, while edge detection pre-processing step in processing pipeline, such contour estimation, object detection , recognition , feature analysis. thresholding rather quick , dirty way see whether or not happening, or extracting out "active" things while edge detection more computer vision related tasks.
instead of explaining how canny edge detection better, i'm going refer literature.
- this page drexel university great thing me started: http://dasl.mem.drexel.edu/alumni/bgreen/www.pages.drexel.edu/_weg22/can_tut.html
- this page computer vision online goes more depth: http://homepages.inf.ed.ac.uk/rbf/hipr2/canny.htm
hope helps!
Comments
Post a Comment