Organisation internationale de normalisation

Summary of discussions

1. m35097 (from image coding)

“Is it worth saying that for either or both of (a) some reference types and (b) some brands, the number space of ItemIDs and TrackIDs is the same, and therefore a track reference box may refer to Items as well as tracks, and an item reference box refer to tracks as well as items?”
Nice idea, but it would have to be under a new brand that establishes this rule. Not yet.
We would like a new edition that covers amendment 1-3 and Cor1-3, in December.

2. m35033 Support for Derived Images in 23008-12

Concerns expressed about the introduction of this technology at DIS stage.

We agree we’d like to be clear here on the direction. We need to decide whether the same item reference type is used to describe ‘this image was derived from these’ (destructive) and ‘the desired image is derived from these by these operations’ (non-destructive). We probably need at least two basic operations in the first version (a) 90 degree rotation and (b) simple cropping (‘clap’). Do we need item reference chains and also a box that holds a single chain? Do we want a simple ‘use a part12 matrix’ operation?

One issue with the ‘chain in a box’ is that the 4CCs of the operations are now a separate registration authority space (they are no longer item types).

We really need to make sure that metadata is only descriptive not prescriptive, so we probably need to take clap and isre out of the ‘descriptive’ area and do something with them (e.g. an image crop operation, and an image composing operation). This contribution suggests that some of isre is actually ‘descriptive’ (e.g. entire width and height) and some ‘prescriptive’.

We like the idea of defining some descriptive data format-independent: width and height, channels, pixel depth (channel depth), etc. (Like a video sample entry).

A revised version of this was uploaded and further revised in details in the meeting to go into the study.

We note that having every operation be an item means it does not have an intrinsic version; should we put in version and/or flags fields as a matter of habit in the more complex editing operations?

3.