An image format for retro-style images
I made this format so that you can make your own image with a hex editor for a retro-ish artstyle. In the future I may need to create an editor to allow making larger images easily.
Each file should start with a header to provide some information about the image.
The first 7 bytes are "IMRETRO". This is the file signature.
The next byte is used for flags for enabled/disabled features used by the decoder.
The first two bits declare the bits-per-pixel of the image. 00 for 1-bit, 01 for
2-bit, and 10 for 8-bit.
Next is a single bit for palette usage: 0 for no palette, 1 to declare that
the file contains a palette. When no palette is present in the file, this means that
the file decoder should choose its own default palette.
The following 2 bits are unused. They are reserved for potential future features.
The sixth and seventh bits are a flag for how many channels each color in the
in-file palette will have. 00 for grayscale (1 channel), 01 for RGB (3
channels), and 10 for RGBA (4 channels). The decoder should ignore this flag
if the in-file palette flag is not set.
The eighth bit is a flag for color accuracy for the in-file palette.
0 for 2 bits per color channel, 1 for 8 bits (a byte) per color
channel. The decoder should ignore this flag if the in-file palette flag is not
set.
Following that are 24 bits for the dimensions: 12 for width and 12 for height. The reason for this limited range of dimensions is to be faithful to the retro-ish goal of this format. In fact, the maximum dimensions are fairly large, but with the goal of supporting at least 480p.
This results in 11 bytes for the header.
The palette will declare the possible colors in the image. The number of colors in your palette depend on the number of bits you chose to use in your header. In 1-Bit mode, you will declare 2 colors, in 2-Bit mode, 4 colors, and in 8-bit mode you will declare 256 colors.
The bits used to declare the palette depends on the color channel and color accuracy flags in the modes.
For example, in 8-bit mode with 4 color channels and the color accuracy flag set (8 bits per channel), the palette would be 8192 bits, or 1024 bytes.
Conversely, in 1-bit mode with 1 color channel and the color accuracy flag
unset (2 bits per channel), only 4 bits would need to be written. Note that this
is less than a full byte. The last byte of the palette should be 0-filled (e.g.
1011 are the colors in 10110000).
After the header comes the actual declaration of the pixels. The number of bits used in each pixel will depend on the "mode" you chose. Only 1 bit for each pixel in 1-Bit mode, 2 bits in 2-Bit mode, etc. Each value for each pixel maps to a color in the palette.
When multiple pixels are stored in a single byte, the bits are used from left to right.
For example, if a byte in 2-bit mode contains the pixels values 1 2 3 and the remaining 2 bits are
unused (this can happen when the number of pixels is not a multiple of 8), then the byte would be
01101100. This style is used so that unused bits can simply be left as 0, and so that the bits
can be read from left to right.
This mode has only two colors: off and on.
The default palette is for "off" to be black and "on" to be white, like you might expect from a Pong console.
This mode has four colors: off, light, strong, and full.
256 colors.
For the nth color, where n is in [0, 256), the RGBA values are n & 3, n >> 2 & 3,
n >> 4 & 3, and n >> 6 & 3. In other words, if n is represented as a byte, then r is the
smallest 2 bits, g is the next 2 bits, etc.
The first 64 colors are completely transparent, but technically have different RGB values.
