Image data format is the internal memory representation of every image. Data formats are described by TImageFormat enumeration member of TImageData structure. Additional information about each of the data formats is stored in TImageFormatInfo structure.

Function name

Usage

GetImageFormatInfo Returns additional information about the given data format GetPixelsSize Returns number of bytes occupied by Width x Height area of pixels in the given data format (works for all data formats) GetFormatName Returns symbolic name of the given data format ConvertImage Converts images between different data formats NewImage Creates new image with the given data format Data Format Categories Image data formats can be divided to these main categories:

ARGB Formats
Each pixel consists of several color channels and optional alpha transparency channel.

Indexed Formats
Each pixel is index to the array of colors called the palette.

Grayscale Formats
Pixels consist of intensity channel and optional alpha transparency channel.

Floating-point ARGB Formats
These formats are similar to ARGB formats but color values are stored as floating-point numbers.

Special Formats
Each special format requires special and unique conversion procedures. Compressed formats like DXTC belong to this category.

Meaning of Data Format Symbolic Names Each data format has its symbolic name. These names are formed in the way which tells something about its internal structure. Symbolic names look like this:

if[(channel1)(channel1sizebits)]..[(channeln)(channelnsizebits)][modifier1]..[modifiern]
For example ifA8R8G8B8 shows that each image pixel contains four 8 bit sized channels. ifR32F format contains one 32 bit channel and F modifier indicates that this is floating-point format. All channels in the symbolic names are listed from left to right, most significant bit (MSB) to least significant bit (LSB). For example, ifA8R8G8B8 is ordered from the MSB channel A (alpha), to the LSB channel B (blue). When traversing image data on little endian machines, the data is stored in memory from LSB to MSB, which means that the channel order in memory is from LSB (blue) to MSB (alpha).
Special data format names do not follow this naming convention. Their symbolic names refer for example to compression technology used (ifDXT5 is data format of image compressed with DXT5).

Symbolic Name Fragment	Info	X	Unused memory space	R	Red color channel	G	Green color channel	B	Blue color channel	A	Alpha transparency channel	Index	Index to palette	Gray	Grayscale intensity channel	F	Indicates that this format stores channel values as floating point numbers

ARGB Formats The following table lists all ARGB formats supported in the current version of Imaging library. Channel values are stored as unsigned integers.

Symbolic Name	Info	ifX5R1G1B1	8 bit format with one bit for each color channel. 5 bits are unused.	ifR3G3B2	8 bit format with 3 bits for read and green and 2 bits for blue channel.	ifR5G6B5	16 bit format with 5 bits for read and blue and 6 bits for green channel.	ifA1R5G5B5	16 bit format with 5 bits for each color and one bit for alpha.	ifA4R4G4B4	16 bit format with 4 bits for each color and alpha.	ifX1R5G5B5	16 bit format with 5 bits for each color and one unused bit.	ifX4R4G4B4	16 bit format with 4 bits for each color and 4 unused bits.	ifR8G8B8	24 bit format with 8 bits for each color.	ifA8R8G8B8	32 bit format with 8 bits for each color and alpha.	ifX8R8G8B8	32 bit format with 8 bits for each color and 8 unused bits.	ifR16G16B16	48 bit format with 16 bits for each color.	ifA16R16G16B16	64 bit format with 16 bits for each color and alpha.	ifB16G16R16	48 bit BGR format with 16 bits for each color.	ifA16B16G16R16	64 bit BGR format with 16 bits for each color and alpha.

Indexed Formats The following table lists all indexed formats supported in the current version of Imaging library. Indexes are stored as unsigned integers. Palette is PPalette32 member of TImageData structure. It is a pointer to array of 32 bit ARGB values in ifA8R8G8B8 format.

Symbolic Name	Info	ifIndex8	8 bit format with 8 bits for index.

Grayscale Formats The following table lists all grayscale formats supported in the current version of Imaging library. Channels values are stored as unsigned integers.

Symbolic Name	Info	ifGray8	8 bit format with 8 bits for grayscale intensity.	ifA8Gray8	16 bit format with 8 bits for alpha transparency and 8 bits for grayscale intensity.	ifGray16	16 bit format with 16 bits for grayscale intensity.	ifGray32	32 bit format with 32 bits for grayscale intensity.	ifGray64	64 bit format with 64 bits for grayscale intensity.	ifA16Gray16	32 bit format with 16 bits for alpha transparency and 16 bits for grayscale intensity.

Floating-point ARGB Formats The following table lists all floating-point formats supported in the current version of Imaging library. Channels values are stored as signed floating-point numbers.

Symbolic Name	Info	ifR32F	32 bit format with one red 32 bit IEEE-754 channel.	ifA32R32G32B32F	128 bit format with 32 bit IEEE-754 channel for each color and alpha.	ifA32B32G32R32F	128 bit BGR format with 32 bit IEEE-754 channel for each color and alpha.	ifR16F	16 bit format with one red 16 bit half-float (s1e5m10) channel.	ifA16R16G16B16F	64 bit format with 16 bit half-float (s1e5m10) channel for each color and alpha.	ifA16B16G16R16F	64 bit BGR format with 16 bit half-float (s1e5m10) channel for each color and alpha.

Special Formats The following table lists all special formats supported in the current version of Imaging library.

Symbolic Name	Info	ifDXT1	Image in this format is compressed with DXT1 compression (15/16 bit interpolated color, zero or one bit alpha).	ifDXT3	Image in this format is compressed with DXT3 compression (16 bit interpolated color, 4 bit explicit alpha).	ifDXT5	Image in this format is compressed with DXT5 compression (16 bit interpolated color, 8 bit interpolated alpha).	ifBTC	Image in this format is compressed with BTC compression (block truncation coding, grayscale, 2bits per pixel).	ifATI1N	Image in this format is compressed using 3Dc+ compression (one channel, 4bits per pixel).	ifATI2N	Image in this format is compressed using original 3Dc compression (two channels, 8bits per pixel).

There maybe some legal issues when using DXTC/S3TC compression (3Dc probably too as it is "subset" of DXT5) as S3 holds patent to this.