It seems that the SHVC specification is missing some information to specify the use of inter-layer motion prediction and sample based prediction for SNR scalability. There exists SamplePredEnabledFlag and MotionPredEnabledFlag that is used to control if resampling of the inter-layer reference picture or its motion field is performed when the resolution of the reference picture and the current picture is different but it does not specify anything specific for the case that the resolutions are identical.

It can be resolved by replacing following text in H.8.1.4

– rsPic is set equal to rlPic.

with

rsPic is derived as follows: – The PicOrderCntVal value of rsPic is set equal to the PicOrderCntVal value of rlPic. – When SamplePredEnabledFlag[ currLayerId ][ rLId ] is equal to 1, the rsPic is set equal to rlPic. – When MotionPredEnabledFlag[ currLayerId ][ rLId ] is equal to 1, the motion field of rsPic is set equal to the motion field of rlPic.

It seems that there is missing some requirements on bitstream conformance when the SamplePredEnableFlag is equal to 1. It can be resolved by following addition in the end of H.8.5:

It is also a requirement of bitstream conformance that the refIdxLX not corresponds to an inter-layer reference picture if SamplePredEnabledFlag [currLayerId][rLId] is equal to 0.

There seem to be missing some text to specify that when inter-layer motion prediction not is used for a certain layer the temporal collocated reference index cannot correspond to an inter-layer reference picture. It can be resolved by the following addition of text in H.8.5:

When 8.5.3.2.7 (Derivation process for temporal luma motion vector prediction) is perfomed it is a requirement of bitstream conformance that the RefPicListX[collocated_ref_idx] not corresponds to an inter-layer reference picture when MotionSamplePredEnabledFlag [currLayerId][rLId] is equal to 0