<div style="line-height:1.7;color:#000000;font-size:14px;font-family:arial"><DIV>At 2013-11-02 06:19:05,"Matt Johnson" <johnso87@illinois.edu> wrote:<BR>>Hi all,<BR>> I'd like to use PredIntraAngs_C() (and other variants of the intra_pred_allangs primitives) for chroma blocks. My understanding from the standard is that neighbor samples for chroma blocks are never filtered (please correct me if I'm wrong!), so the neigbor pixel arrays provided via the pAbove1 and pLeft1 arguments should never be used. The logic in PredIntraAngs_C() decides whether to use the filtered or unfiltered neighbor pixels using a LUT indexed by block size and angular direction; the decision does not use the bLuma argument. When I pass in bLuma == false, pAbove1 and pLeft1 are still used in the underlying PredIntraAngBufRef().<BR>> Should I:<BR>>1) pass in pAbove1 == pAbove0 and pLeft1 == pLeft0 to circumvent the issue?<BR>You can do it so.</DIV>
<DIV> </DIV>
<DIV>>2) patch PredIntraAngs_C() to decide appropriately based on bLuma?<BR>The PredIntraAngs_C() design to get all of 33 modes Intra prediction, I think it can't be work for Chroma</DIV>
<DIV>The input bLuma will pass to PredIntraAngBufRef()<BR></DIV></div>