I have a matix of class double but when i am converting it to an image by imwrite it's class changes to uint8 but i want it to be double type what should i do?

matix to image

Answer 1

[Walter Roberson]

https://www.mathworks.com/matlabcentral/answers/376954-is-it-possible-to-write-a-double-data-matrix-into-tiff-images-movie#answer_300040

https://www.mathworks.com/matlabcentral/answers/333229-how-can-i-save-an-image-and-read-it-again-without-change-its-value#comment_443711

There is no supported way to write double precision images in matlab . There are supported ways to write single precision tiff and dicom. Only tiff and png and dicom support double precision at all in their standards but matlab does not support creating them (unless possibly tiff support was added recently ) . One of the links above leads to aa File Exchange contribution that claims to support double precision tiff.

In my experience everyone who has asked for imwrite to support double precision has had a weak understanding of image file formats . The people who have had good reason to deal with double precision images have directly asked about the tiff class or about dicom transfer syntaxes.

[sb icse]

So what should I do when I want a double class matrix to be converted to an image of double class?Can you help me out with the code.

[Walter Roberson]

if you had followed the links II provided you would have found

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

But do you really have more 24 bits of floating point precision needed for your data, and a viewer that can handle double precision images (ImageJ might but most software will not ) ?

Or are you just trying to do lsb steganography with dft or dct or wavelet and trying to ensure that you can do the inverse after storing as an image?

[Walter Roberson]

People who have a serious need for 64 bit images almost always have a context that controls what kind of image is to be used , such as a need to be able to exchange data with an electron microscopy application that requires TIFF and a particular set of tags.

[sb icse]

j=1143 1142 1134 1132 1119 1101 1104 1099 1085

this are the pixel values present in the matrix j.Now I want it to be converted to an image but when I do so all the values are changing to 255.

[Guillaume]

@Walter, PNG does not support floating point formats. It only supports integer up to 16 bits. See Table 11.1 of the specs for which bit depth is supported for each colour format (greyscale, rgb, indexed)

[Walter Roberson]

As I wrote in one of the posts i linked to:

To write floating point values into PNG files, you need to store them as raw blocks of uninterpreted data, just binary blobs; image display programs would not be able to display the result. If I recall correctly, MATLAB does not provide a way to writing these kinds of blocks, but I am not certain of that.

[Walter Roberson]

png can act as aa container with

blobs of nearly arbitrary size . The same is true of tiff but is not true of jpg

[Guillaume]

Ah, yes sure, you can write FP data in private chunks of a png file. in fact, I did just that a long time ago as an extension to a program I used to use that created images in single precision. Saved the 32 bit FP raw image in a private chunk andd an 8-bit integer version in the IDAT chunk, but I wouldn't say that PNG support FP images out of the box, the way TIFF does.

[Guillaume]

And, yes I don't think matlab lets you access private chunks of a PNG. Perhaps, they ought to provide a PNG interface similar to the TIFF interface. Or since, libpng is shipped with matlab, you could always directly interact with it.

[Walter Roberson]

Hence my "at all".

Though looking again with jpeg II guess you could use aa private exif tag which would get you 4 bytes worth of chunk length . (matlab provides no way to write exif. there is aa fex contribution to call an external program to write exif)

Answer 2

[Image Analyst]

Save it as a .mat file with save(). Any problem with that? What app is going to read these in after you write it out? If it's just MATLAB again, simply use save() and load().