[Python]Manipulating histogram values

asked 2018-03-15 23:25:48 -0500

Hi! Well, i'm fairly new to Python but i have some coding background on some other languages like C. I could have gone for C++ but i wanted to learn Python so here i am, treading on unknown territory and i hope you guys can help me : )

The thing is, i've got an image:

SourceImage

And i was able to extract (most of) its background, giving me my region of interest:

ThresholdedImage

Now, i wan't to access the HSV values of the pixels in my RoI, which i did using:

hist_H = cv2.calcHist([img],[0],mask,[256],[0,256]) #For the H channel

And so on for the S and V channels (altough i'm just gonna use the H and S).

I need to get the average or predominant value of each specific channel, but using NumPy's mean/median functions compute the number of pixels, and not the value of those pixels (which are between 0 and 256). If i use the mean/median directly on the channel (say, the H channel) then it gives me the mean/median of all pixel values, but unfortunately it will compute the black pixels from my mask :/

I need this data so I can determine HSV ranges for ripe/unripe fruits and the like.

Does anyone have any idea of what could be done in this case?

edit retag flag offensive close merge delete

Comments

"I need this data so I can determine HSV ranges for ripe/unripe fruits" -- hmm, you already got a nice mask, and the histogram. isn't that all you need for a comparison ?

imho, it's much better, to compare histograms for this (longer features, more precision), than hsv ranges.

you could go as far as training an SVM on histograms of ripe/unripe apples, and let it predict() the outcome furtheron.

berak gravatar imageberak ( 2018-03-17 10:38:31 -0500 )edit

Hi, Well, i didn't think about using the whole array of values from the histogram to train something. I thought it would be way easier to train something based on one variable, but i can understand how better it would be if i could train it to see the whole picture (hehe). I will try finding some sources on the subject of SVMs. Thanks!

Navarrox gravatar imageNavarrox ( 2018-03-18 14:04:57 -0500 )edit