3

I want to obtain the pixel values for RGB from image.I did it in Matlab and Python but i obtained different values especially at the green value. I'll appreciate if you have an advice about this matter. Thanks Here is my code in python

from PIL import Image
import numpy as np

im = Image.open("lena.jpg")
imPixelVal = np.ones(np.size(im))
imSize = np.size(im)

for i in range (0,imSize[0]):
            for j in range (0,imSize[1]):         
                ij = i , j
                p = im.getpixel(ij)
                imPixelVal[i,j] = (0.2989 * p[0]) + (0.5870 * p[1]) + (0.1140 * p[2])
                print p[0]
                print p[1] 
                print p[2]

Also this is the code in Matlab:

Im=imread('lena.jpg');
Img = (ones(size(Im,1),size(Im,2))); 

for i=1:size(Im,1)
      for j=1:size(Im,2)
          Img(i,j)=0.2989*Im(i,j,1)+0.5870*Im(i,j,2)+0.1140*Im(i,j,3);
      end
end
Im(1,1,1)
Im(1,1,2)
Im(1,1,3)
lisa
  • 31
  • 3
  • 3
    Can you provide a [MCVE](http://stackoverflow.com/help/mcve) and an example of the data differences? – Ffisegydd Aug 22 '14 at 08:43
  • it might be interesting to know which data format your images are having. – OBu Aug 22 '14 at 08:49
  • Give specifics on what code you used and what the differences are. – Joop Aug 22 '14 at 08:54
  • I editted and added the code for Python and Matlab – lisa Aug 22 '14 at 11:52
  • Since JPEG is a lossy format, different readers can be off by +/-1 from each other quite easily depending on how they round the decoded values. Schorsch's answer is likely to be the correct one though. – Mark Ransom Aug 22 '14 at 19:48
  • @Mark That's interesting to know. However, I checked for this case and all values are identical between Python and Matlab – Schorsch Aug 23 '14 at 15:02

1 Answers1

3

It appears that the images are read in a different "direction" in Python compared to Matlab.
If you change your python code to: 

ij = j , i

instead of 

ij = i , j

you'll get the same output as in Matlab.

If you want Matlab to give the same results as Python, you'd have to flip i and j there: 

Img(j,i)=0.2989*Im(j,i,1)+0.5870*Im(j,i,2)+0.1140*Im(j,i,3);

Here's how I figured this out through simple debugging:

  • First, I got the image from here and saved it as .jpg.

  • Then, I changed the Matlab loops to 

    for i=1:2
    for j=1:2

    So that I would only get the first 4 pixels.

  • By printing both i, j and the contents of Im I got: 

    i = 1, j = 1
225, 137, 125

i = 1, j = 2
227, 139, 127

i = 2, j = 1
224, 136, 124

i = 2, j = 2
226, 138, 126

  • Now, I did the same in python: 

    for i in range (0,2):
    for j in range (0,2):

  • This gave me: 

    (0, 0)
225 137 125
(0, 1)
224 136 124
(1, 0)
227 139 127
(1, 1)
226 138 126

  • This showed me that the order is different between Matlab and Python.

  • Hence, changing i and j in Python from ij = i, j to ij = j, i will reproduce the Matlab results.
Schorsch
  • 7,761
  • 6
  • 39
  • 65