ISO IEC 23001-16:2021 pdf download – Information technology — MPEG systems technologies — Part 16: Derived visual tracks in the ISO base media file format.
If a referenced track is a member of an alternate group or switch group, or if the reference is to a track group, then the reader should pick a track from the group as the input to the derived visual track. NOTE 2 The TrackSelectionBox can be used to provide guidance on the selection between members of an alternate group or switch group. Similarly, if a referenced image item is a member of an alternate group (which may contain both tracks and images), then the reader should pick one member of the group as the input to the derived visual track. A derived sample contains an ordered list of the derivation operations to be performed, each derivation operation applying a derivation transformation on an ordered list of inputs. The layer syntax element in TrackHeaderBox has no impact on ordering the inputs for derived samples. The four-character codes of derivation transformation from all derivation operations used by the derived samples in the track are listed in the DerivedVisualSampleEntry , and also default inputs and parameter values can be supplied there. A derived sample in the track may use all or some of the derivation operations listed in the linked DerivedVisualSampleEntry , but derived samples shall not use a derivation operation not listed in the sample entry. The derived sample durations document the time over which the derivation represented by the ordered list of derivation operations is active. Therefore, the number of samples defined in a derived visual track does not necessarily match 1:1 with the number of input image items or samples of input tracks that are being transformed. A single derivation duration may span multiple samples in the source track(s), and also derivation transformations in derived samples may have ‘internal time structure’ (e.g. a cross- fade) so the picture may change during the sample duration. This is in contrast to ‘classic video’. Derived visual tracks do not respect edit lists on inputs. They operate on the composition timeline (i.e. before the application of edit lists) of their input tracks (including on derived visual tracks when used as visual inputs). However, the input tracks shall not have edit lists. Any edit lists of the input tracks shall be ignored if present. NOTE 3 When time-alignment adjustment between input tracks is needed, signed composition offsets in input tracks can be used. NOTE 4 A derived visual track can have an edit list; thus, a derived visual track using the identity transform, and with an edit-list, can provide a visual output that is a temporal re-mapping of the input track. The inputs for a derivation operation in a derived sample can be either input image items from file-level MetaBox or intervals (possibly spanning multiple samples) of input video tracks, image sequence tracks, metadata items or metadata tracks, the visual output of a preceding derivation operation or a default input fill picture. Transformative item properties or transformations (e.g. clean aperture, track matrix etc …) associated with input image items or samples of input tracks are always applied before performing the derivation operation. NOTE 5 If a derived sample needs to refer to one explicit sample value in a referred track (other than the time- aligned sample value), an item can be created and referred to that has the same data as the desired sample value. The visual inputs in a derived sample shall have consistent pixel aspect ratio and bit depth. The input image items, samples of input tracks or derived samples may have various width and height. When differences in width and height result in pixels that never get ‘painted ’ by a derivation operation, those empty pixels are filled according to the value of default_derivation_input parameter signalled in DerivedVisualTrackConfigRecord (black, white or grey pixels). When differences in width and height result in pixels that end up outside the visual output size by a derivation operation, those pixels are cropped. This default behaviour may be overridden by derivation operation specifications. A derived sample is reconstructed by performing the specified derivation operations in sequence. Some derivation operations can be marked as non-essential which indicates that the derivation operation may be skipped by the reader. However, the operations marked as essential shall be used in order to obtain a valid derived sample.