Edit pixel values in an image as numpy array

Question

I am reading an image like so:

img = np.array(Image.open(test_dir + image_name))

What I want to do is find a region in the array where the value is high (250 or more) and reduce it by 10:

rows = img.shape[0]
cols = img.shape[1]
pixel = []
for x in range(0,rows):
    for y in range(0, cols):
        if x >= 122 and x <= 160 and y >= 34  and y <= 71:
            if img[x,y]>= 250:
                img[x,y] = img[x,y] -10
                pixel.append(img[x,y])

So the region I am looking at according the the unaltered image should be a box from (122, 34) to (160,71), and it should have some pixels over 250, but somehow when I run this code I end up with nothing in the pixel list

lightalchemist · Accepted Answer · 2018-09-13 02:32:07Z

1

Is img a 3 dimensional array? If so, then your test img[x, y] >= 250 is testing if an array of 3 elements is greater than 250. This will result in an error and cause your script to terminate at that point. You did not say if this is happening.

If it is a 3 dimensional array and you want to check if all channels have value greater than 250 at position x, y, then you have to check for np.all(img[x, y] >= 250) instead of img[x, y] >= 250.

In general you want to avoid loops and use vectorized operations to speed things up.

min_row = 122
max_row = 161
min_col = 34
max_col = 72
threshold = 250

row_slice = slice(min_row, max_row)
col_slice = slice(min_col, max_col)
roi = img[row_slice, col_slice] >= threshold
img[row_slice, col_slice][roi] -= 10

or more succinctly but less readable,

roi = img[min_row:max_row, min_col:max_col] >= threshold
img[min_row:max_row, min_col:max_col][roi] -= 10

edited Sep 13, 2018 at 2:32

answered Sep 13, 2018 at 1:49

lightalchemist

10.2k4 gold badges49 silver badges55 bronze badges

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

Comments

Mark Setchell · Accepted Answer · 2018-09-13 08:36:15Z

The simplest way is one line:

im[im>250]-=10

Demonstration

Start with this 8x8 gradient image:

Here it is enlarged:

Then use IPython like this:

# Load image as L (greyscale)
im = np.array(Image.open('image.png').convert('L'))

# View contents
im
Out[16]: 
array([[255, 255, 255, 255, 255, 255, 255, 255],
       [219, 219, 219, 219, 219, 219, 219, 219],
       [182, 182, 182, 182, 182, 182, 182, 182],
       [146, 146, 146, 146, 146, 146, 146, 146],
       [109, 109, 109, 109, 109, 109, 109, 109],
       [ 73,  73,  73,  73,  73,  73,  73,  73],
       [ 36,  36,  36,  36,  36,  36,  36,  36],
       [  0,   0,   0,   0,   0,   0,   0,   0]], dtype=uint8)

# Do required processing
im[im>250]-=10

# Review contents
In [18]: im
Out[18]: 
array([[245, 245, 245, 245, 245, 245, 245, 245],
       [219, 219, 219, 219, 219, 219, 219, 219],
       [182, 182, 182, 182, 182, 182, 182, 182],
       [146, 146, 146, 146, 146, 146, 146, 146],
       [109, 109, 109, 109, 109, 109, 109, 109],
       [ 73,  73,  73,  73,  73,  73,  73,  73],
       [ 36,  36,  36,  36,  36,  36,  36,  36],
       [  0,   0,   0,   0,   0,   0,   0,   0]], dtype=uint8)

Let's try decreasing even more pixels, just for fun:

In [19]: im[im>100]-=10

In [20]: im
Out[20]: 
array([[235, 235, 235, 235, 235, 235, 235, 235],
       [209, 209, 209, 209, 209, 209, 209, 209],
       [172, 172, 172, 172, 172, 172, 172, 172],
       [136, 136, 136, 136, 136, 136, 136, 136],
       [ 99,  99,  99,  99,  99,  99,  99,  99],
       [ 73,  73,  73,  73,  73,  73,  73,  73],
       [ 36,  36,  36,  36,  36,  36,  36,  36],
       [  0,   0,   0,   0,   0,   0,   0,   0]], dtype=uint8)

kevinkayaks · Accepted Answer · 2018-09-13 06:59:46Z

0

Provided your image is grayscale this will do it.

inds = np.where(img>=250)
inds = inds[(inds.T[0]>=122)&(inds.T[0]<=160)&(inds.T[1]>=34)&(inds.T[1]<=71)]
img[inds]-=10

edited Sep 13, 2018 at 6:59

answered Sep 13, 2018 at 1:20

kevinkayaks

2,7461 gold badge16 silver badges30 bronze badges

Comments

Hans Lehnert · Accepted Answer · 2018-09-13 19:22:38Z

0

For a more compact solution you can try

roi = img[122:161, 34:72]
pixel = roi[roi >= 250]
roi[roi >= 250] -= 10

edited Sep 13, 2018 at 19:22

answered Sep 12, 2018 at 22:03

Hans Lehnert

4744 silver badges10 bronze badges

5 Comments

Benjamin Over a year ago

Neither roi or pixel are used in the final line... Please review.

Mark Setchell Over a year ago

@Benjamin Actually, I just looked at this more carefully and, you are correct, the first 2 lines are not needed, but... the final line works on its own as it is the same as my answer which I demonstrated to work in my answer.

Hans Lehnert Over a year ago

@Benjamin You are right about roi, I've fixed that, thanks. As for, the second line, it isn't explicitly asked but the code in the original question also extracts the values of the pixels in the variable pixel

Hans Lehnert Over a year ago

@MarkSetchell The first line is actually needed since the OP just wants to search within a certain area.

Benjamin Over a year ago

There has been an edit here that makes my comment ambiguous... and pixel is still not being used.

Collectives™ on Stack Overflow

Edit pixel values in an image as numpy array

4 Answers 4

Comments

Comments

Comments

5 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

4 Answers 4

Comments

Comments

Comments

5 Comments

Your Answer

Sign up or log in

Post as a guest

Related