PIL rotate image colors (BGR -> RGB)

Question

I have an image where the colors are BGR. How can I transform my PIL image to swap the B and R elements of each pixel in an efficient manner?

Solution 1

Assuming no alpha band, isn't it as simple as this?

b, g, r = im.split()
im = Image.merge("RGB", (r, g, b))

Edit:

Hmm... It seems PIL has a few bugs in this regard... im.split() doesn't seem to work with recent versions of PIL (1.1.7). It may (?) still work with 1.1.6, though...

Solution 2

I know it's an old question, but I had the same problem and solved it with:

img = img[:,:,::-1]

Solution 3

Just to add a more up to date answer:

With the new cv2 interface images loaded are now numpy arrays automatically.
But openCV cv2.imread() loads images as BGR while numpy.imread() loads them as RGB.

The easiest way to convert is to use openCV cvtColor.

import cv2
srcBGR = cv2.imread("sample.png")
destRGB = cv2.cvtColor(srcBGR, cv2.COLOR_BGR2RGB)

Solution 4

Adding a solution using the ellipsis

image = image[...,::-1]

In this case, the ellipsis ... is equivalent to :,: while ::-1 inverts the order of the last dimension (channels).

Solution 5

This was my best answer. This does, by the way, work with Alpha too.

from PIL import Image
import numpy as np
import sys 

sub = Image.open(sys.argv[1])
sub = sub.convert("RGBA")
data = np.array(sub) 
red, green, blue, alpha = data.T 
data = np.array([blue, green, red, alpha])
data = data.transpose()
sub = Image.fromarray(data)

View more solutions

Comments

[Claudiu]

the solution was to load the image properly in the first place =P but this seems lovely

[Claudiu]

Ah useful, thanks. Note you can suggest an edit and you get +2 rep if it gets approved

[user2692263]

thanks, Claudiu, I'll try that: """Edits must be at least 6 characters; is there something else to improve in this post?""" Well - that did not go so well ;-)

[Zach Garner]

Pretty much the only reason you would have a BGR image is if you're using OpenCV. Your solution is the right thing to do. The other top answer is functional, but would be slow in processing large images.

[Luke Yeager]

np.roll will convert BGR to RBG, not RGB. If you want to do this in numpy, you can use data[...,[2,1,0]] to swap the channels. But if you're already using OpenCV or PIL, just go with Martin Beckett's response.

[Peter9192]

The data has 3 dimensions: width, height, and color. ::-1 effectively reverses the order of the colors. The width and height are not affected.

[Wesley Ranger]

Good, I got it following your note and my testing.::-1 is actually a shorthand of numpy operation [start:end:step], and start/end is decided automatically.

[DollarAkshay]

Why is this not on top ?

[Ryan Soklaski]

Because the simplest solution is literally just: img = img[..., ::-1]

[Martin Beckett]

Does cv2 guarantee that images are contiguous? OpencCV (at least in c++) will start each new row on a 4byte boundary so you have to be careful moving raw bytes around

[Eliethesaiyan]

thanks for this...i am using them in the same codes,got me crazy for a while@RyanSoklaski that is actually more confusing without knowing which is default for opencv or PIL but very neat

[EuWern]

for clarity, img = img[:, :, : :-1] is equivalent to img = img[:, :, [2,1,0]]. I think the later is better as it is more explicit.

[Ryan Soklaski]

img[:, :, : :-1] is not equivalent to img[:, :, [2,1,0]]. The former utilizes basic index and the latter uses advanced indexing. The latter will make a copy of your image while the former does not.

[Vinícius M]

This is also significantly more performant than the cv2.cvtColor function suggested in an answer above.

[hoohoo-b]

Another numpy alternative to @LukeYeager's suggestion: np.flip(image_bgr, axis=2)

[Lewis Morris]

This is how I do it usually, Although if its got alpha Image.fromarray(np.array(img)[:,:,:3][:,:,::-1])

[Muhammad Yasirroni]

Can you elaborate your answer? Because obviously you need to use np.shape or someting like that to get the dimension. Also what is bgr_buf?