How invert only one colour in RGB image?

Hello!

I want to change background in my rgb image from black to white. How I can do it without inversing others colours?

[Walter Roberson]

First step: figure out how to reliably determine whether a particular pixel is part of the background. Is the background always exactly all 0,0,0? (If it is a jpeg image it is highly likely that some background pixels are not 0,0,0)

[Pavel M]

it is png image and for the beginning let the background is exactly 0,0,0

[Pavel M]

Ok, i have written this code, it works

img=imread('image.png');
[x,y,~]=size(img);
for i=1:x
    for j=1:y
        if(img(i,j,1) == 0 && img(i,j,2) == 0 && img(i,j,3) == 0)
            img(i,j,1) = 254;
            img(i,j,2) = 254;
            img(i,j,3) = 254;
        end
    end
end
imshow(img);

but image has become illegible and blurry in some places. How it can be optimized?

[Walter Roberson]

Optimized version, no loop:

img(repmat(all(img==0,3),1,1,3)) = 254;

but image has become illegible and blurry in some places

It probably had anti-aliasing done at some point: if it was ever a jpeg then this is almost certain.

[Image Analyst]

@Pavel M Attach your image if you want any more help. Unless your image is a pure, synthesized computer graphics image, inverting the "background" will likely leave edge artifacts where your background touches your foreground. What is the real purpose or need to do what you said you wanted to do?

[Pavel M]

For the beginning i wanted to invert black to white.

For the second, after inverting i want to do numbers on scales x and y more clear

[Pavel M]

It is ideal variant

Answer 1

[DGM]

The easy way to do this is L inversion. You can do this the hard way, or you can just use MIMT tools:

inpict = imread('scope.png');
outpict = imtweak(inpict,'hsl',[0 1 -1]); % invert L
imshow(outpict)

Of course yellow on white is hard to read, so desaturating might help.

inpict = imread('scope.png');
outpict = imtweak(inpict,'hsl',[0 0.7 -1]); % invert L, reduce S
imshow(outpict)

Otherwise, you could always try to mask off certain parts of the image and adjust them independently.

Imtweak() is part of MIMT, and is available at the link above.

[Pavel M]

Super! Thx!

What function can mask off certain parts in your MIMT tools?

[DGM]

Honestly, for this, I'd probably just use IPT images.roi.Polygon() to create polygonal selections and then use createMask() on them. It really depends what the goals are and whether it needs to be repeated on multiple images. If it's only a one-off task, it'd probably be easier to do the image work in an image editor.

[Pavel M]

Ok, and could you say how do not invert gray colours in your code above?

[DGM]

I'm assuming you're after the floating toolbar things. This is one example. I can't really provide an example of the mask generation, since I did that by manual selection. I attached the mask image.

inpict = imread('scope.png');

% mask created by manual selection using immask() in HSV
mk = imread('scopemask.png');
mk = imfill(mk,'holes'); % fill in holes

% adjust the whole image
outpict = imtweak(inpict,'hsl',[0 0.7 -1]); % invert L, reduce S

% but use the mask to revert the selected areas
outpict = replacepixels(inpict,outpict,mk);

imshow(outpict)

[Pavel M]

Nice! Thank you