[x265] two questions about x265 code

[Steve Borho]

On Dec 4, 2013, at 5:19 AM, zhs <2594450788 at qq.com> wrote:

> Hi, steve
>  
> Sorry to bother you:
>  
> I found your email in the x265 source code and I imagine you must be familiar with the code base.  I just started looking at the code base and there are a few places that aren't clear to me.  If you could, could you please enlighten me on the following ?

Hello,

Please send questions to x265-devel at videolan.org so everyone can benefit from the Q/A, added to cc:

> 1. As I understand it, CTURow::processCU() decides what to do with grammar elements, and actual encoding takes place in FrameEncoder::encodeSlice.  The two principal operations in CTURow::processCU() are m_cuCoder.compressCU(cu) and m_cuCoder.encodeCU(cu).  I don't understand the purpose of the latter encodeCU(cu) operation.

This is an unfortunate left-over from the HM, we would like to combine the two passes.  The first pass analyses each CU and decides on the QuadTrees and the second pass generates the final bitstream.

> 2. TComYuv::removeHighFreq() is named as if it's a lowpass filter, but what it does is computes dst[x]  = (dst[x]  << 1) - src[x] .  I don't understand what doing this accomplishes or how it relates to low-pass filtering.

This is also a hold-over from the HM; I thought I had removed this function.  It was used in their bidir motion search but we no longer use it.

--
Steve Borho
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 842 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20131204/5c9f3cae/attachment-0001.sig>