.JPG File Format – Application segments

The common segments

JFIF

JFIF (ITU-T T.871) defines the first segment of many files, directly after SOI. It says the components are Y, Cb and Cr (or Y alone), with the transform shown on the coding page.

A JFXX segment (JFXX\0, then 1 byte) holds a thumbnail in another form: 10 a whole JPEG, 11 width, height, a 768-byte palette and 1-byte pixels, 13 width, height and RGB pixels.

Exif

After Exif\0\0 the body is a TIFF file: a byte-order header followed by image file directories holding the camera's details, the date, the exposure, the location and usually a thumbnail. Offsets in it count from the start of the TIFF header, not from the start of the file. Tags are described in the TIFF reference and the Exif standard.

XMP

After the identifier comes an XMP packet of XML, up to one segment long. A larger packet is split: the main segment holds the common part, and xmp/extension segments hold the rest, each with the packet's 32-character GUID, the total size (4 bytes), this piece's offset (4 bytes), and the bytes.

ICC profiles

A colour profile is split across APP2 segments, each with ICC_PROFILE\0, a 1-based piece number (1 byte), the number of pieces (1 byte), and the piece. The pieces joined in order are the profile.

Multi-Picture Format

Some cameras store several images in one file, one after another. The first image's APP2 MPF\0 segment, in a TIFF layout like Exif's, lists each image's type, size and offset; the offsets count from the segment's TIFF header. The extra images sit after the first image's EOI, and each is a whole JPEG file.

Photoshop and Adobe

APP13 holds Photoshop's image resources: each is 8BIM, a 2-byte ID, a padded name, a 4-byte size, and data padded to an even length. A resource with ID 0404 holds IPTC text.

Comments

A COM segment (FF FE) holds text of any kind, of no meaning to decoding. T.81 does not say what encoding it uses.