fimg_file_format_v2.0.adoc

FIMG v2.0

FIMG is a float image file format.

File suffix: .fimg

Status

2022-05-01

Version 2.0 is a specification work in progress…​
This version of fimg will probably not be ready and see the day of light as it takes a big step from keeping things as simple as v1.0.

Simplicity was more or less the main reason for the fimg file format. If you want complicated stuff then there are probably other file formats out there to make it complicated for you…​.

File format description

Simple enough:

Table 1. Image file header

Name	Data format	Description
Magic number	4 byte (32 bit in total)	The ASCII sequence FIMG for "Float IMaGe". 0x46, 0x49, 0x4D, 0x47
Major version	1 unsigned byte (8bit)	Current version value 2.
Minor version	1 unsigned byte (8bit)	Current version value 0.
Image width	32bit big endian unsigned integer	in pixels
Image height	32bit big endian unsigned integer	in pixels
Amount color channels	1 byte	Usually 1 (for grayscale), 3 (for RGB), or 4 (for RGBA or CMYK)
Channel data stream configuration	1 byte	0 = the channel data for each pixel is interleaved for each pixel in the data stream. I.e. example RGB image (R,G,B,R,G,B,R,G,B,…​,R,G,B) 1 = the channel data for each channel is stored complete channel chunks in the data stream after another. I.e. example RGB image (R,R,R,…​,R,G,G,G,…​,G,B,B,B,…​,B)

Table 2. Channel header. One or more blocks of channel headers (as many as specified by Amount color channels). The order of the channel headers is important.

Name	Data format	Description
Channel name	8 byte (64 bit in total)	Channel name or code in ASCII. Reserved values are: RGB_R, RGB_G, RGB_B for RGB red, green, and blue channels. ALPHA is reserved for alpha (transparency) channel. MONO is reserved for monochrome (grayscale) image information. CMYK_C, CMYK_Y, CMYK_M, and CMYK_K is reserved for CMYK images. YCbCr_Y, YCbCr_Cb, and YCbCr_Cr is reserved for YCbCr images. Unused trailing bytes of channel name are set to zero/null/0x00.
Value factor	32bit big endian float	Value to multiply to the every channel value read. If not applicable or used it is set to 1.0.
Value offset	32bit big endian float	Value to add to the every channel value read (after factor multiplication). If not applicable or used it is set to 0.0.
Lowest value	32bit big endian float	The lowest value in the channel. This value is as the values are stored in the channel data and is subject to be calculated according to Value factor and Value offset.
Highest value	32bit big endian float	The highest value in the channel. This value is as the values are stored in the channel data and is subject to be calculated according to Value factor and Value offset.
Lowest possible value	32bit big endian float	The lowest possible value in the channel. Should be equal or less than Lowest value field. Used to set the 0% intensity or black level value. This value is as the values are stored in the channel data and is subject to be calculated according to Value factor and Value offset.
Highest possible value	32bit big endian float	The highest possible value in the channel. Should be equal or higher than Highest value field. Used to set the 100% intensity level value. This value is as the values are stored in the channel data and is subject to be calculated according to Value factor and Value offset.
Channel data offset	32bit big endian unsigned integer	The byte offset of the actual channel data after the end of Image header data (see above). The channel data offset, in the storage configuration as interleaved per pixel, all point to the first data byte in the interleaved stream. (An additional offset of (channel index * 4 byte) will have to be used if reading separate channel information.

Table 3. Image pixel data, pixel by pixel starting at top left and then from left to right, top line to bottom line.

Data format	Value
32bit big endian float	R value of RGB. Value range from 0.0 and up (no limit).
32bit big endian float	G value of RGB. Value range from 0.0 and up (no limit).
32bit big endian float	B value of RGB. Value range from 0.0 and up (no limit).