processing opencv image pixels quickly using python interface

Question

Using python-interface for OpenCV, one can easily access the pixel of image using [] operator,like this:

img = cv.LoadImage('test.jpg')
pixel = img[10,10]

variable pixel here is a python tuple object like (10,20,30) (3 channels for example), it's not very convenient to process computation since tuple type does not support operator '-' or '+'. If I hope to do a substruction on a pixel like 255 - (10,20,30), I must code like this:

import numpy as np
pixel = tuple( np.array([255,255,255]) - np.array(pixel) )

is there a faster and easier solution?
Another qustion : is there a way to make a substraction on all pixels, like using a matrix substraction in Matlab: 255 - img (don't use OpenCV build-in function).

jfs · Accepted Answer · 2011-04-26 04:13:20Z

3

You could use cv2array()/array2cv() functions from adaptors.py in the opencv source distribution and perform all your computation using numpy arrays. 255 - imgarr works in this case. Example (stripped-down version of cv2array() for read-only arrays):

assert isinstance(img, cv.iplimage) and img.depth == cv.IPL_DEPTH_8U
a = np.frombuffer(img.tostring(), dtype=np.uint8)
a.shape = img.height, img.width, img.nChannels
print 255 - a

edited Apr 26, 2011 at 4:13

answered Apr 23, 2011 at 21:00

jfs

417k210 gold badges1k silver badges1.7k bronze badges

Sign up to request clarification or add additional context in comments.

5 Comments

PinkyJie Over a year ago

my opencv version is 2.1, I found adaptors.py in \OpenCV2.1\interfaces\swig\python, and there was not a function called cv2array() or array2cv()

jfs Over a year ago

@PinkyJie: google.com/…

jfs Over a year ago

@PinkyJie: On the other hand the above method might be outdated. Does your version have cv.fromarray() function? What do you get on numpy.asarray(img)? opencv.willowgarage.com/documentation/python/…

PinkyJie Over a year ago

I installed opencv using "OpenCV-2.1.0-win32-vs2008.exe", a pre-compiled edition, so no cv.fromarray() function, and numpy.asarray(img) return a iplimage object array, not numpy.ndarray.

jfs Over a year ago

@PinkyJie: I've added example code in case np.asarray() doesn't return anything useful.

Collectives™ on Stack Overflow

processing opencv image pixels quickly using python interface

1 Answer 1

5 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

5 Comments

Your Answer

Sign up or log in

Post as a guest

Related