Formation of a 16-bit PGM file

Question

C++ / C#

I decided to present them in the form of graphic files for experiments with surfaces (many thanks for the idea Eddy_Em . To do this, I took an excellent simple PGM file format. Format description . In general, everything works fine with the 8-bit format:

#include <stdio.h>
#define width 255
#define height 255
int main()
{
  unsigned int i,j;
  FILE *img_file = NULL;
  img_file = fopen("dat/greyscale.pgm", "w");
  char tmp=0;
  fprintf(img_file, "P5 \n%d %d \n255 \n", width, height);
  for (i = 0; i < width; i++) {
    for (j = 0; j < height; j++) {
      tmp=i;
      fwrite(&tmp, sizeof(char), 1, img_file);
    }
  }
  if (img_file != NULL) 	fclose(img_file);
  return 0;
}

The result is excellent. The size of the output file is 65,043 bytes



But it's worth trying to switch to a two-byte format, everything goes wrong.
I replace the data variable with short

unsigned short tmp=0;

In the file header I say that we now have 16 significant bits:

fprintf(img_file, "P5 \n%d %d \n65535 \n", width, height);

And now I'm writing short

fwrite(&tmp, sizeof(unsigned short), 1, img_file);

Full version of the code

#include <stdio.h>

#define width 255
#define height 255
int main()
{
  unsigned int i,j;
  FILE *img_file = NULL;
  img_file = fopen("dat/greyscale.pgm", "w");
  unsigned short tmp=0;
  fprintf(img_file, "P5 \n%d %d \n65535 \n", width, height);
  for (i = 0; i < width; i++) {
    for (j = 0; j < height; j++) {
      tmp=i;
      fwrite(&tmp, sizeof(unsigned short), 1, img_file);
    }
  }
  if (img_file != NULL) 	fclose(img_file);
  return 0;
}

I get an erroneous file, 130,070 bytes in size, which looks like a black square when opened. The HEX editor shows that the result is correct.

Even if the byte order is not correct, the output should still not be a black square, but some kind of overflow.

What to do and who is to blame? How do I generate a 16-bit grayscale file? Can use some other uncompressed format (if so, which one?).

UPD The problem was solved as follows, the output to the file was implemented strongly-typed

{
       tmp=(char)(i&0xFF);
       fwrite(&tmp, sizeof(char), 1, img_file);
       tmp=(char)((i>>8)&0xFF);
       fwrite(&tmp, sizeof(char), 1, img_file);
}

Graphic editors show the resulting file in black, but you can already work with it and the FFT is done ...

Answer 1

If you want to store accurate data in a graphic file, I strongly advise you to use the FITS format!
An excellent description + examples of working with cfitsio are on the Internet. Well, I myself immodestly can offer to read my LiveJournal on this subject.
If you don't like FITS, you can choose tiff (perhaps even with compression) or png. I worked with both formats: everything is fine, there are no difficulties.