These utility routines look at the headers of JPEG and PNG files to find the pixel dimensions of any images supplied by the user for cover art and figures.

• §1. JPEG files
• §2. PNG files

§1. JPEG files. The following code, contributed by Toby Nelson, either finds the pixel width and height of a given JPEG file and returns TRUE or, if it can't read the file or doesn't recognise the header as having JPEG format, returns FALSE.

JPEG is properly speaking not a file format but a compression technique: the routine below works with either JIF (JPEG Interchange Format) or its simpler cousin JFIF (JPEG File Interchange Format).

We scan the file looking for "markers", each of which begins with an 0xFF byte and is followed by a marker-type byte which is neither 0x00 nor 0xFF. The compulsory marker SOI must appear at the start of the file, providing one way to detect probable JPEGs by looking at the first two bytes. There must also eventually be a start of frame marker, for the actual image: this can have many forms, but in all cases tells us the height and width.

int ImageFiles::get_JPEG_dimensions(FILE *JPEG_file, unsigned int *width, unsigned int *height) {
    unsigned int sig, length;
    int marker;

    if (!BinaryFiles::read_int16(JPEG_file, &sig)) return FALSE;
    if (sig != 0xFFD8) return FALSE;  0xFF (marker) then 0xD8 (SOI)

    do {
        do {
            marker = getc(JPEG_file);
            if (marker == EOF) return FALSE;
        } while (marker != 0xff);  skip to next 0xFF byte

        do {
            marker = getc(JPEG_file);
        } while (marker == 0xff);  skip to next non FF byte

        if (!BinaryFiles::read_int16(JPEG_file, &length)) return FALSE;  length of marker

        switch(marker) {
             all variant forms of "start of frame": e.g., 0xC0 is a baseline DCT image
            case 0xc0:
            case 0xc1: case 0xc2: case 0xc3:
            case 0xc5: case 0xc6: case 0xc7:
            case 0xc9: case 0xca: case 0xcb:
            case 0xcd: case 0xce: case 0xcf: {

                 fortunately these markers all then open with the same format
                if (getc(JPEG_file) == EOF) return FALSE;  skip 1 byte of data precision

                if (!BinaryFiles::read_int16(JPEG_file, height)) return FALSE;
                if (!BinaryFiles::read_int16(JPEG_file, width)) return FALSE;

                return TRUE;
            }
            default:
                if (fseek(JPEG_file, (long) (length - 2), SEEK_CUR) != 0) return FALSE;  skip rest of marker
        }
    }
    while (marker != EOF);

    return FALSE;
}

§2. PNG files. The PNG file must start with a signature which indicates that the remainder contains a single PNG image, consisting of a series of chunks beginning with an IHDR chunk and ending with an IEND chunk ("Portable Network Graphics (PNG) Specification", 2nd edition, section 5.2). We only need to scan the IHDR chunk, of which the pixel width and height are the first two words (section 11.2.2).

int ImageFiles::get_PNG_dimensions(FILE *PNG_file, unsigned int *width, unsigned int *height) {
    unsigned int sig1, sig2, length, type;

     Check PNG signature
    if (!BinaryFiles::read_int32(PNG_file, &sig1)) return FALSE;
    if (!BinaryFiles::read_int32(PNG_file, &sig2)) return FALSE;
    if ((sig1 != 0x89504e47) || (sig2 != 0x0d0a1a0a)) return FALSE;

     Read first chunk
    if (!BinaryFiles::read_int32(PNG_file, &length)) return FALSE;
    if (!BinaryFiles::read_int32(PNG_file, &type)) return FALSE;

     First chunk must be IHDR
    if (type != 0x49484452) return FALSE;

     Width and height follow
    if (!BinaryFiles::read_int32(PNG_file, width)) return FALSE;
    if (!BinaryFiles::read_int32(PNG_file, height)) return FALSE;
    return TRUE;
}