Create image from unsigned char buffer

Question

I'm capturing fingerprints using a device called Secugen Pro 20, it has its own SDK for Linux, and i want to capture the fingerprint image and save it as any image format.

They have this typedef unsigned char BYTE;

I declared my imageBuffer

BYTE *CurrentImageBuffer;

Then i allocate memory to it using the devices specs

CurrentImageBuffer = malloc(device_info.ImageWidth*device_info.ImageHeight);

And at some point at my code i capture image and pass CurrentImageBuffer as argument to the capture function:

SGFPM_GetImageEx(m_hFPM, CurrentImageBuffer, GET_IMAGE_TIMEOUT, NULL, GET_IMAGE_DESIRED_QUALITY)

Thats what the variable looks right after this line of code ( i can confirm that it captured a finger):

I just don't understand how to proceed creating an image from this buffer, as it doesn't look like a ByteArray

I don't even know if thats the right place to get my image from, but that looks like the right place because its a buffer, right?.

OBS: I'm new to C

Answer 1

This is a small sample program to write an 8-bit graylevel image into a Windows BMP file:

#include <stdio.h>

typedef unsigned char Byte;

int writeBMPGray8(FILE *f, int w, int h, const Byte *data)
{
  unsigned bytesPerRow = (w + 3) & ~3; // align to 4 bytes (requirement)
  unsigned size
    = 14 // Bitmap file header size
    + 12 // DIB header size
    + 256 * 3; // palette size
  unsigned gap = size;
  size = (size + 3) & ~3; // align to 4 bytes (requirement)
  gap = size - gap; // get size of gap between end of headers and raw data
  unsigned offs = size; // store offset of raw data
  size += h * bytesPerRow; // bitmap data size in file
  /* write Bitmap file header (14 bytes) */      
  { const Byte buffer[14] = {
      'B', 'M', // magic code
      size & 0xff, size >> 8 & 0xff, size >> 16 & 0xff, size >> 24 & 0xff, // size of BMP file in bytes
      0, 0, // reserved
      0, 0, // reserved
      offs & 0xff, offs >> 8 & 0xff, offs >> 16 & 0xff, offs >> 24 & 0xff // starting offset of pixel data
    };
    if (fwrite(buffer, sizeof buffer, 1, f) != 1) return -1; // ERROR!
  }
  /* write DIB header (12 bytes) */
  { const Byte buffer[12] = {
      12, 0, 0, 0, // size of this header
      w & 0xff, w >> 8 & 0xff, // bitmap width in pixels
      h & 0xff, h >> 8 & 0xff, // bitmap height in pixels
      1, 0, // number of color planes, must be 1
      8, 0 // number of bits per pixel
    };
    if (fwrite(buffer, sizeof buffer, 1, f) != 1) return -1; // ERROR!
  }
  /* write color palette (3 * 256 bytes) */
  for (int i = 0; i < 256; ++i) { // make a gray level palette
    Byte buffer[3] = { i, i, i };
    if (fwrite(buffer, sizeof buffer, 1, f) != 1) return -1; // ERROR!   
  }
  /* write gap (to align start address of raw data with 4 */
  for (int i = 0; i < gap; ++i) {
    if (fputc(0, f) < 0) return -1; // ERROR!
  }
  /* write raw data */
  for (int y = 0; y < h; ++y) { // for all rows
    int x = 0;
    for (; x < w; ++x) { // for all columns
      if (fputc(*data++, f) < 0) return -1; // ERROR!
    }
    // write row padding
    for (; x < bytesPerRow; ++x) {
      if (fputc(0, f) < 0) return -1; // ERROR!
    }
  }
  /* done */
  return 0;
}

int main()
{
  /* a sample image 6 x 8, gray level */
  enum { w = 6, h = 8 };
  const Byte imgRaw[w * h] = {
    0x00, 0x30, 0x60, 0x90, 0xc0, 0xf0,
    0x02, 0x32, 0x62, 0x92, 0xc2, 0xf2,
    0x04, 0x34, 0x64, 0x94, 0xc4, 0xf4,
    0x06, 0x36, 0x66, 0x96, 0xc6, 0xf6,
    0x08, 0x38, 0x68, 0x98, 0xc8, 0xf8,
    0x0a, 0x3a, 0x6a, 0x9a, 0xca, 0xfa,
    0x0c, 0x3c, 0x6c, 0x9c, 0xcc, 0xfc,
    0x0e, 0x3e, 0x6e, 0x9e, 0xce, 0xfe
  };
  FILE *f = fopen("test.bmp", "wb");
  if (!f) return 1; // ERROR!
  if (writeBMPGray8(f, w, h, imgRaw)) return 1; // ERROR!
  if (fclose(f)) return 1; // ERROR!
  return 0; // success
}

The sample image provides some kind of gradients horizontally and vertically. I've chosen a width of 6 intentionally to check/show that row alignment is done properly.

The implementation is based on the description in Wikipedia BMP file format.

To keep it short, I encoded the simplest format – the ancient BITMAPCOREHEADER of Windows 2.0 and OS/2 1.x. (MS Paint can load this as well as the Windows 10 preview. I tested with GIMP which loaded as well without any complaints.)

This is how it looks in GIMP:

Answer 2

The easiest way to get an image is to make a NetPBM PGM image - see Wikipedia NetPBM page.

So, if your image is say 640 px wide by 480 px tall, you would get a buffer from your SDK with 307,200 bytes and you would write that to a file and check it has the correct length. Call that image.raw.

Now you just need a PGM header, and as your image is greyscale and binary, you need a P5 header.

So, in Terminal you can put a header on:

{ printf "P5\n640 480\n255\n" ; cat image.raw ; } > image.pgm

If you are unfamiliar with that syntax, you can get the same with:

printf "P5\n640 480\n255\n" >  image.pgm
cat image.raw               >> image.pgm

And you can view that image with feh, gimp, Photoshop etc.

---

If you want to make it into a BMP, or JPEG, or PNG, use ImageMagick which is installed on most Linux distros and is available for macOS and Windows:

magick image.pgm image.png

or

magick image.pgm image.jpg

If your version of ImageMagick is v6 or older, use convert in place of magick:

convert image.pgm image.png

Answer 3

If you have correctly captured the image in CurrentImageBuffer, you can write this as raw file using the code fragment below:

        fp = fopen(rawFileName,"wb");
        fwrite (CurrentImageBuffer, sizeof (BYTE) , device_info.ImageHeight*device_info.ImageWidth , fp);
        fclose(fp);

As I have used the same environment, I am sending the above fragment from my working codebase. Actually, raw file is later converted to template which is later used for matching / identification and not directly used for viewing etc. Variable rawFileName stores the name of the file as char array (string) where this buffer is stored.