How to save a matix as an image?

I have a matrix finalMat of data type double with values in the range of 0 to 255. It actually corresponds to some image. I am displaying it as below :

imshow(finalMat, []);

But when i try to save it using the below code, the image saved is completly white.

imwrite(finalMat,'myImage.jpeg','JPEG');

I want to save the image on disk without changing the values in the finalMat matrix. When I read the saved image i.e myImage.jpeg, I must get the same values as in finalMat. Can somebody please help in saving the image? Thank you in advance..

[Adam]

The help for imwrite states:

Note: imwrite converts indexed images to RGB before writing data to JPEG files, because the JPEG format does not support indexed images.

So I would suggest saving to some other format if you just want 0 to 255 raw data saved.

[Devadath Prabhu]

Thank you. Can you please suggest me which format can be preferred to save the raw data in matrix?

[Adam]

I'm not an expert on different image formats, I literally just took a quick look at the Matlab help for imwrite and saw that. You can easily experiment with different file formats though.

[Devadath Prabhu]

Sure! Thank you :)

Answer 1

[Image Analyst]

To save a floating point image as a TIFF file, this should do it.

% Create floating point image.
rgbImage = rand (10, 20, 3);
% Image must be single precision.
rgbImage = single(rgbImage);
% Display it.
imshow(rgbImage, 'InitialMagnification', 1000)
axis('on', 'image');
% Create tiff object.
fileName = '_floatingPointImage.tif';
tiffObject = Tiff(fileName, 'w')
% Set tags.
tagstruct.ImageLength = size(rgbImage,1); 
tagstruct.ImageWidth = size(rgbImage,2);
tagstruct.Compression = Tiff.Compression.None;
tagstruct.SampleFormat = Tiff.SampleFormat.IEEEFP;
tagstruct.Photometric = Tiff.Photometric.MinIsBlack;
tagstruct.BitsPerSample = 32;
tagstruct.SamplesPerPixel = size(rgbImage,3);
tagstruct.PlanarConfiguration = Tiff.PlanarConfiguration.Chunky;
tiffObject.setTag(tagstruct);
% Write the array to disk.
tiffObject.write(rgbImage);
tiffObject.close;
% Recall image.
m2 = imread(fileName)
% Check that it's the same as what we wrote out.
maxDiff = max(max(m2-rgbImage))

[Walter Roberson]

https://www.mathworks.com/matlabcentral/answers/7184-how-can-i-write-32-bit-floating-point-tifs-with-nans#answer_9845

https://www.mathworks.com/matlabcentral/fileexchange/30519-export-image-to-tif-or-tiff-file-of-selected-data-type

[Image Analyst]

OK thanks Wlater for showing me how to correct the errors above. I corrected that code so this below is the same as above. This works now in Windows 10:

% Create floating point image.
rgbImage = rand (10, 20, 3);
% Image must be single precision.
rgbImage = single(rgbImage);
% Display it.
imshow(rgbImage, 'InitialMagnification', 1000)
axis('on', 'image');
% Create tiff object.
fileName = '_floatingPointImage.tif';
tiffObject = Tiff(fileName, 'w')
% Set tags.
tagstruct.ImageLength = size(rgbImage,1); 
tagstruct.ImageWidth = size(rgbImage,2);
tagstruct.Compression = Tiff.Compression.None;
tagstruct.SampleFormat = Tiff.SampleFormat.IEEEFP;
tagstruct.Photometric = Tiff.Photometric.MinIsBlack;
tagstruct.BitsPerSample = 32;
tagstruct.SamplesPerPixel = size(rgbImage,3);
tagstruct.PlanarConfiguration = Tiff.PlanarConfiguration.Chunky;
tiffObject.setTag(tagstruct);
% Write the array to disk.
tiffObject.write(rgbImage);
tiffObject.close;
% Recall image.
m2 = imread(fileName)
% Check that it's the same as what we wrote out.
maxDiff = max(max(m2-rgbImage)) % Should be zero.

[Federico Fioretti]

thank you both very much. I have one last question. The matrix that I have to save as .tiff file, originally was an image that I had uploaded to matlab as a .mat file, then I modified it so that its values ​​were floating point, and now, at this point, I have to extract this matrix as a .tiff file. This to say that the matrix I have is a double data type instead of an rgb image. Do you think the code yow wrote will work the same?

I'm sorry for the questions, but I recently started using matlab.

Thank you very much!

[Walter Roberson]

Yes, the code referenced will write floating-point data as-is no matter what range it is (though be careful about single or double precision)

However, what the receiving program does with the double precision values is a different question.

In my experience, most of the time if you double() integer values, and do not take care later to return them to integer data type, then the flow is probably broken.

[Linh Nguyen]

@Federico Fioretti matrix A

if you want to save matrix A as a image but A show a white image.

imwrite(A/max(A(:))

[Federico Fioretti]

@Image Analyst I'm sorry but the code doesn't work. I obtain a white image except for a small part of it. I think it's beacuse the code considers only values between 0 and 1, while I have values also higher than 1 and i could also have negative values. So i need a bigger range of values to be considered. How can i do it?

[Image Analyst]

@Federico Fioretti it does work. Look here is proof:

% Create floating point image.
rgbImage = 3000 * rand (10, 20, 3);
% Image must be single precision.
rgbImage = single(rgbImage);
% Display it.
imshow(rgbImage, 'InitialMagnification', 1000)
axis('on', 'image');
% Create tiff object.
fileName = '_floatingPointImage.tif';
tiffObject = Tiff(fileName, 'w')
% Set tags.
tagstruct.ImageLength = size(rgbImage,1); 
tagstruct.ImageWidth = size(rgbImage,2);
tagstruct.Compression = Tiff.Compression.None;
tagstruct.SampleFormat = Tiff.SampleFormat.IEEEFP;
tagstruct.Photometric = Tiff.Photometric.MinIsBlack;
tagstruct.BitsPerSample = 32;
tagstruct.SamplesPerPixel = size(rgbImage,3);
tagstruct.PlanarConfiguration = Tiff.PlanarConfiguration.Chunky;
tiffObject.setTag(tagstruct);
% Write the array to disk.
tiffObject.write(rgbImage);
tiffObject.close;
% Recall image.
m2 = imread(fileName)
% Check that it's the same as what we wrote out.
maxDiff = max(max(m2-rgbImage)) % Should be zero.

You can see I have values as high as 3000 and I get back those same values because maxDiff is zero. Don't try to display this in MATLAB because MATLAB expects all RGB images to be in the range of 0-1. I thought this was going to be used in your photogrammetry program so don't worry about how it's displayed in MATLAB. If you really want to see it in MATLAB you'll have to divide by the max value to get a scaled image before you call imshow().

scaledRGB = rescale(rgbImage, 0, 1);
imshow(scaledRGB);

Answer 2

[Image Analyst]

If your image is uint8, I recommend PNG. It's lossless compression. About a third the size of BMP and TIFF images. More than JPEG but with perfect quality, unlike the crummy images you can get from jpeg if you compress too much. Don't use jpeg for image analysis. Create your filename with a .png extension and you don't need to use the third argument for imwrite.

imwrite(rgbImage, 'fubar.png');

You can find out what other formats are allowed with the imformats command:

>> imformats

EXT       ISA     INFO        READ      WRITE     ALPHA  DESCRIPTION                     
-----------------------------------------------------------------------------------------
bmp       isbmp   imbmpinfo   readbmp   writebmp  0      Windows Bitmap
cur       iscur   imcurinfo   readcur             1      Windows Cursor resources
fts fits  isfits  imfitsinfo  readfits            0      Flexible Image Transport System
gif       isgif   imgifinfo   readgif   writegif  0      Graphics Interchange Format
hdf       ishdf   imhdfinfo   readhdf   writehdf  0      Hierarchical Data Format
ico       isico   imicoinfo   readico             1      Windows Icon resources
j2c j2k   isjp2   imjp2info   readjp2   writej2c  0      JPEG 2000 (raw codestream)
jp2       isjp2   imjp2info   readjp2   writejp2  0      JPEG 2000 (Part 1)
jpf jpx   isjp2   imjp2info   readjp2             0      JPEG 2000 (Part 2)
jpg jpeg  isjpg   imjpginfo   readjpg   writejpg  0      Joint Photographic Experts Group
pbm       ispbm   impnminfo   readpnm   writepnm  0      Portable Bitmap
pcx       ispcx   impcxinfo   readpcx   writepcx  0      Windows Paintbrush
pgm       ispgm   impnminfo   readpnm   writepnm  0      Portable Graymap
png       ispng   impnginfo   readpng   writepng  1      Portable Network Graphics
pnm       ispnm   impnminfo   readpnm   writepnm  0      Portable Any Map
ppm       isppm   impnminfo   readpnm   writepnm  0      Portable Pixmap
ras       isras   imrasinfo   readras   writeras  1      Sun Raster
tif tiff  istif   imtifinfo   readtif   writetif  0      Tagged Image File Format
xwd       isxwd   imxwdinfo   readxwd   writexwd  0      X Window Dump

[Devadath Prabhu]

My image matrix is of type double. After I save in PNG format, when I read it back, I am not getting the same matrix back. I need the same values back when I load the saved image..

[Image Analyst]

If you have a double image, use save() to save it as a mat file. Most of the standard image formats are for integer valued images.

[Devadath Prabhu]

Okay. I will convert the double matrix to integer type. Now how do I save it as image and read it such that I get back the same integer values?

[Walter Roberson]

If you use imwrite() to write uint8 or uint16 images, you will get back the same result when you read them unless you used an image format that uses lossy compression. JPG image format uses lossy compression and so should be avoided for this purpose. PNG and TIFF do not use lossy compression and so should be able to give you back exactly what you wrote.

If you use uint32 or uint64 then you might need to use special parameters or special ways to write the image, as not many image formats support those.

[Devadath Prabhu]

Thanks a lot! This helped :)

[Walter Roberson]

If you really want to write floating point numbers in an image format, then I suggest using the Tiff class, or DICOM format.

If your purpose is steganography then you are going about it the wrong way.

[Image Analyst]

How? For tiff, this doesn't work:

grayImage = imread('cameraman.tif');
grayImage = double(grayImage) + rand(size(grayImage));
imwrite(grayImage, '_deleteme.tif');
recalledImage = imread('_deleteme.tif'); % It's uint8, not double

and for dicom, the documentation says this for the input image.

Data Types: int8 | int16 | uint8 | uint16

Answer 3

[Guillaume]

Matlab consider the type of the matrix to establish what intensity range corresponds to fully dark-fully white.

If the matrix is of type double, then matlab uses the range 0-1 to represent intensities. Any value above 1 is considered the same as 1 and is maximum brightness. Hence your intensities 1 to 255 are all considered the same = white.

However, if the matrix is of type uint8, then matlab uses the range 0-255, just as you have. So convert your matrix to uint8:

imwrite(uint8(finalMat) 'myImage.jpeg');

[Devadath Prabhu]

Thanks for the answer. When saved using this, if I read the saved image using below code, the contents in readIm is different from that of finalMat.

readIm=imread('myImage.jpeg');

I want the values of finalMat to be saved as image and when I read that image, I want those values back. Please help.

[Guillaume]

It is different in what way?

I assume that the values in finalMat are integers. If not, then there's no way you can save it as a jpg file without losing the decimals.

I also assume that your matrix is an 3D matrix (of size m x n x 3). If it's only a 2D matrix, you'll have to convert it to rgb before writing to jpg:

imwrite(repmat(uint8(finalMat), [1 1 3]), 'myImage.jpeg')

And to get your original image, just keep one of the page.

readIm = imread('myImage.jpeg');
readIm = readIm(:, :, 1);

[Devadath Prabhu]

Matrices are different in the sense that their values are different though the image looks similar. But I want the values also to be same. In my case, both finalMat and readIm are 512*1024. The first 8*8 block of finalMat has these values -

131.087903712260  126.588085931488  129.019131897930  127.986820756743  128.036797886933  128.129163153756  128.901568706468  127.913467179027
126.865651056741  128.644726154519  126.696839938151  129.320745973049  129.064319024373  127.759203600434  128.006382827703  128.032312069063
128.547542279462  128.013326895391  128.389098782354  128.799496025384  128.055922974060  127.763243851185  127.922268272279  128.470671480691
128.028037000488  127.433327438777  127.479136613125  128.980352019314  128.193475727530  128.480245823339  128.413770202099  127.880514709290
129.227871666373  128.063934629160  127.465928877913  128.144312598059  127.757980122585  129.250265336802  128.413391078785  128.787456156361
127.965837250422  126.915520053310  127.628369773152  127.390251727192  127.971102024689  127.836604160943  128.531568211385  127.799176238100
128.197089880864  128.309110329146  128.403008073977  127.913385806887  128.332535509389  127.758264955073  127.422464120998  127.273960679635
128.041637714197  128.369029343605  127.807345577358  127.854815737404  128.197686086408  128.134149584508  128.051525804888  127.880870927027

After trying the code you suggested, first 8*8 block of readIm has these values -

128  128  128  128  128  128  128  128
128  128  128  128  128  128  128  128
128  128  128  128  128  128  128  128
128  128  128  128  128  128  128  128
128  128  128  128  128  128  128  128
128  128  128  128  128  128  128  128
128  128  128  128  128  128  128  128
128  128  128  128  128  128  128  128

I want readIm to have same values as in finalMat. How can I do that?

Answer 4

[Image Analyst]

Save it as a .mat file instead of an image format file

save('my floating point image.mat', 'myImage'); % Save myImage into a .mat file.

To recall it later:

s = load('my floating point image.mat'); % Load data in .mat file into a structure.
myImage = s.myImage; % Extract the floating point image from the structure.

[Federico Fioretti]

I try to explain better what my problem is. I have a matrix of noninteger values and i have to save this matrix as an image because I have to use it later in an other software.

I have written a code that loads a file.mat in matlab, modifies it, and then i have to export this file (that is the matrix of non integers values) as an image format file. Can you help me?

Answer 5

[Milind patil]

Brother, Simply use .png extension instead of .jpg.

imwrite(finalMat,'myImage.png'); 

This will give u exact values for sure.

[Image Analyst]

It will NOT. Just try this:

m = 255 * rand(512);         % Floating point in range 0-255.
imwrite(m, 'm.png');         % Save PNG file
mRecalled = imread('m.png'); % uint8 integers
isequal(m, mRecalled)        % Returns false
delete('m.png')              % Delete temporary test file.

and you will see.

Note from Devadath's comment that the numbers are floating point. If they were integers then what you say is true, but not in the floating point case.

[DEEPAK DAIYA]

use

imwrite(mat2gray(finalMat),'myImage.jpeg','JPEG');

[Image Analyst]

DEEPAK:

That also does not work:

finalMat = 255 * rand(512);         % Floating point in range 0-255.
imwrite(mat2gray(finalMat),'myImage.jpeg','JPEG');
mRecalled = imread('myImage.jpeg'); % uint8 integers
whos mRecalled
isequal(m, mRecalled)   % Returns false
delete('myImage.jpeg')  % Delete temporary test file.

See, the recalled image is still uint8, not double, thus the values will change if there are any fractional values and it's not all integers.

[GAB_image]

Following the intention to save a floating point matrix as an image, I have tried first to normalize a TIF image (orinignally unit 8 as when is loaded with imread() funciton ) to have its values between 0 and 1.

However, when I try to write back the normalized matrix (a floating point matrix) to a TIF file image, it is saved agin as a unit 8 image (as the original one).

How could it be possible to save the foating potin matrix to a new TIF file image keeping the values between 0 and 1 obtained from the normalizartion of the original image and not get it saved back to unit 8?

% read image
im=imread('imge.tiff');
% convert to double
im = double(im);
% normalize between 0 and 1
im1 = (im-min(im(:)))/(max(im(:))-min(im(:)));
% save the normalized ijmage as TIF 
imwrite(im1,'image_normalized_0_1.tiff');
% However when I relaod it is still the original unit 8 image not the normalized one:
im2  = imread('image_normalized_0_1.tiff');

[Image Analyst]

To keep it as floating point, save it as a .mat file

% Save floating point image:
save('image_normalized_0_1.mat', 'im1');
% Recall floating point image:
s = load('image_normalized_0_1.mat');
im1 = s.image_normalized_0_1;

[Yogesh Bhattarai]

last code results- Reference to non-existent field 'image_normalized_0_1'. ??

[Nini]

@Image Analyst, If im1 is the matrix ( with floating point values) , can you please explain how to save it as an image and then read it with the exact same values.

'.mat' doesn't give us an image right?

[Image Analyst]

.mat does give the correct values, and it's an array, which means it's an image.

grayImage = imread('cameraman.tif');
grayImage = double(grayImage) + rand(size(grayImage));
save('_deleteme.mat', 'grayImage');
% Now read it back in to see if we get the same values.
s = load('_deleteme.mat')
recalledImage = s.grayImage % A double, as required.

[Linh Nguyen]

I have a same problem, I want to get a .png file from complex double matrix.

[Linh Nguyen]

I want a .png file.

I have a double matrix.

img = imread('cameraman.tif'); img = double(img);

if I use:

imwrite(A, 'file_name.png');      

it is different from imshow(A, []);

[Walter Roberson]

@Linh Nguyen

img = imread('cameraman.tif');
A = rescale(img);
imwrite(A, 'file_name.png');

imshow(img, []); title('original')

imshow(A); title('rescaled')

Answer 6

[Linh Nguyen]

@Milind patil give me your email - address, I want to ask about this topic. Do you find any way to solve it?

[Federico Fioretti]

@Linh Nguyen Hi, I have your same problem. Have you found any way to solve it?

[Image Analyst]

Save it as a .mat file instead of an image format file

save('my floating point image.mat', 'myImage'); % Save myImage into a .mat file.

To recall it later:

s = load('my floating point image.mat'); % Load data in .mat file into a structure.
myImage = s.myImage; % Extract the floating point image from the structure.

[Federico Fioretti]

Thanks for the answer. I explain better my problem. I have a matrix with non-integers elements. I need to save this matrix in an image format (jpg or png) without losing the values, beacause I have to use this image later in an other photogrammetry sofrtware. So the .mat file is useless. Can you help me?

[Federico Fioretti]

The matrix is a double matrix of non-integers values.

[Image Analyst]

Tell me ALL the file formats your photogrammetry program can import.

[Federico Fioretti]

.jpg .png .bmp .tiff for sure. The others I have to see in the PC i have in the office so now i can't.

also .dng and .raw files

[Walter Roberson]

There is a File Exchange contribution that can write floating-point TIFF files.

Caution: it is not common for programs to support floating-point TIFF files. If I recall correctly, ImageJ does support them, but Windows Media does not; I am not sure about Adobe programs.

[Image Analyst]

To save a floating point image as a TIFF file, this should do it.

% Create floating point image.
rgbImage = rand (10, 20, 3);
% Image must be single precision.
rgbImage = single(rgbImage);
% Display it.
imshow(rgbImage, 'InitialMagnification', 1000)
axis('on', 'image');
% Create tiff object.
fileName = '_floatingPointImage.tif';
tiffObject = Tiff(fileName, 'w')
% Set tags.
tagstruct.ImageLength = size(rgbImage,1); 
tagstruct.ImageWidth = size(rgbImage,2);
tagstruct.Compression = Tiff.Compression.None;
tagstruct.SampleFormat = Tiff.SampleFormat.IEEEFP;
tagstruct.Photometric = Tiff.Photometric.MinIsBlack;
tagstruct.BitsPerSample = 32;
tagstruct.SamplesPerPixel = size(rgbImage,3);
tagstruct.PlanarConfiguration = Tiff.PlanarConfiguration.Chunky;
tiffObject.setTag(tagstruct);
% Write the array to disk.
tiffObject.write(rgbImage);
tiffObject.close;
% Recall image.
m2 = imread(fileName)
% Check that it's the same as what we wrote out.
maxDiff = max(max(m2-rgbImage))