convert gray image back to rgb

how to back grat to rgb gray2rgb function not found

[Benoit Espinola]

What do you mean by "back to rgb"?

Do you mean to put original colors back into the image? If so, that is impossible, you lost the information when saving the image in black and white.

If you mean "RGB data format" then you just need to repeat the grey value 3 times such as: RGB = [G G G] where G is a column vector.

[AliHdr]

I have done this but i can not use imwrite anymore

there is and error

<<Data with 9 components not supported for JPEG files.>>

how can I overwrite new rgb file with old gray file?

[Image Analyst]

@AliHdr evidently JPG can't store multispectral or volumetric images. Either use a mat file or save each slice as its own image.

Answer 1

[Image Analyst]

If you have a color map - a mapping of what intensity should go to what color, like what imread() or rgb2ind() can give you - then you can use ind2rgb.

rgbImage = ind2rgb(grayImage, colormap);

Otherwise use

rgbImage = cat(3, grayImage, grayImage, grayImage);

grayImage needs to be in the range 0-255 uint8 if you want to display it.

[Image Analyst]

Start a new question and attach your image and segmentation code.

[Alexandar]

What do you mean by create a colormap and how can you do it within the scales that you'd like?

[Image Analyst]

@Alexandar There are a bunch of built-in colormap functions such as hsv, turbo, and jet. Or you can make up your own 256 by 3 matrix with values in the range 0-1.

To apply the colormap to the data range you want, you can use the clim or caxis() function.

Answer 2

[Shaun VanWeelden]

Alternatively

rgb=img(:,:,[1 1 1]); does the same thing as Jan's, but is a little shorter maybe and easier to remember. img is your image you want to convert to rgb obviously

[su mon aung]

rgb=img(:,:,[1 1 1]; this equation still remains my image as grayscale image

[Walter Roberson]

It will become an RGB image whose colors all happen to be gray.

My tests show that if you take all possible RGB values and convert them to gray, that on average 65536 different combinations map to each value. 7 combinations map to complete black and 7 map to complete white; there are four shades that are created by 111642 different combinations. Clearly it is not possible to take a grayscale image and map it back to "the" original color.

[SHAIKH TAUSEEF HASAN]

rgb=img(:,:,[1 2 3]);

[Image Analyst]

No, that throws an error if img is a gray scale image since there is no third dimension for gray scale images.

Answer 3

[Jan]

Asking Google for "Matlab gray2rgb" would be a good idea.

But a general method is:

RGB = cat(3, Gray, Gray, Gray);

[Muhammad Waqas]

it can work some time

[Walter Roberson]

No, this always works for gray images.

What it does not work for is

• pseudocolor images: use ind2rgb for those
• black and white images represented as datatype logical and for which the converted image is to be displayed using imshow() instead of image(): for those either use image() or use double() on the result of the cat()

[john]

what do u mean by use double? its giving me error when i use imshow

[Walter Roberson]

@john

sample_logical_2d = rand(64,80) > 0.8;
imshow(sample_logical_2d)

sample_logical_3d = cat(3, sample_logical_2d, sample_logical_2d, sample_logical_2d );
try
    imshow(sample_logical_3d)
catch ME
    fprintf('oooo! imshow did not like 3D logical!')
    disp(ME)
end

oooo! imshow did not like 3D logical!

  MException with properties:

    identifier: 'images:imageDisplayValidateParams:expected2D'
       message: 'If input is logical (binary), it must be two-dimensional.'
         cause: {}
         stack: [7×1 struct]
    Correction: []

sample_double_3d = double(sample_logical_3d);
imshow(sample_double_3d)

Answer 4

[shehbaz Ali]

This is not possible to to convert black and white image to gray image. Because you don't what will be colors of any pixel for rgb picture.

[Shaun VanWeelden]

black and white to grayscale is easy, gray=uint8(bw.*254+1), and to convert that to rgb just do one of the other answers

[Walter Roberson]

gray = uint8(bw*255);

or

 gray = double(bw);

[DGM]

I think it's fairly obvious what @shehbaz Ali meant, and I would agree. The conversion from a grayscale or color image to a logical mask irreversibly discards information.

For more details, see the explanation written on this page:

[Walter Roberson]

You cannot gain information when you convert black and white to grayscale, or grayscale to color -- but you can create the grayscale or color image with the same amount of information.

[DGM]

That's a good way to put it. :)