c++ - Unable to find the suitable masks to calculate bad pixel percentage for disparity map -


i working on stereo vision problem , need calculate bad pixel percentage. have depth map ground truth of cones image, need calculate bad pixel percentage 3 categories. given in middlebury site :

bad pixel categories.

as seen in screen-shot, here 3 masks: non-occ, , disc. how generate images each of them. tried threshold function edge detection function. results far desired. know formula bad pixel percentage. first have generate non occlude image, discontinuity image,etc. not able figure out image processing generate of images...

the input image(ground truth) :

groundtruth.

these results want achieve...

the desired results are:

for non-occluded :

non-occ

and discontinuity image :

disc

and including occluded, non-occluded regions in image :

all

looking @ images, intuitively used thresholding , edge detection methods generate non occluded region images , results far ideal.

i think not right way.

which masks 1 one must apply obtain results?

any appreciated.


Comments

Popular posts from this blog

angular - Ionic slides - dynamically add slides before and after -

minify - Minimizing css files -

Add a dynamic header in angular 2 http provider -