[Cin] DAR, PAR and SAR, again

[Terje J. Hanssen]

Den 23.01.2024 18:35, skrev Andrea paz via Cin:
> I did the following test:
>
> NOTE: I use SAR as pixel aspect ratio i.e. SAR= Sample aspect ratio. I
> use PAR as frame aspect ratio i.e. PAR= Picture aspect ratio. I use
> PAR=W/H.
>
> I took the following 4 videos:
>
> Source_1:
> bd.m2ts; 1440x1080; SAR 1:1; DAR 4:3
> Compositor: Display stretched vertically; square px.
>
> Source_2:
> 5sec_dv01_03.dv; 720x576; SAR 16:15; DAR 4:3
> Compositor: Display corrected; px nearly square.
>
> Source_3:
> 5sec_hdv01_04.m2t.ts; 1440x1080; SAR 4:3; DAR 16:9
> Compositor: Correct display; rectangular px.
>
> Source_4:
> bd_2.m2ts; 1920x1080; SAR 1:1; DAR 16:9;
> Compositor: Correct display; square px.

Just a short comment to the above:

If Source_1 results from HDV 1440x1080 (check the video file with 
ffprobe), I think it should have the same parameters as
Source_3:
5sec_hdv01_04.m2t.ts; 1440x1080; *SAR 4:3; DAR 16:9

*
>
> I created a project (empty) 1440x1080; DAR=16:9  [SAR can be
> calculated from: SAR = 16:9 / 4:3 = 1.3333 i.e. 4:3]
> I imported 4 source files into Resources window.
> What I notice in the Compositor is:
>
> source_1 (same WxH; different SAR; different DAR) ==> non-square
> Pixels (px); correct display; no crop/pad.
>           Since WxH are the same the display remains correct. But since
> the DARs are different then the aspect ratio changes adjusting to that
> of the project (from 4:3 to 16:9). As a result the pixels enlarge
> (similarly, adjusting to the design SAR: from 1:1 to 4:3), but without
> crop/pad because they do not change the No. of pixels, but simply
> those that are there extend horizontally.
> source_2 (different WxH; different SAR; different DAR) ==> non-square
> Pixels; flattened display; pad.
>           The worst case because all aspect ratios are different
> between source and project. Changing from smaller WxH_source to larger
> WxH_project results in a pad i.e., an addition of side black bands.
> Going from DAR 4:3 to 16:9 also results in a change in aspect ratio.
> As in the case of source_1 the pixels stretch horizontally to change
> from 16:15 to 4:3. The black sidebands arise precisely from the
> extension of pixels that is not sufficent to fill the DAR_project; the
> missing pixels are black and in such numbers as to fill the rest of
> the project format.
>
> source_3 (same WxH; same SAR; same DAR) ==> non-square Pixels; correct
> display; no crop/pad.
>           The best case since all aspect ratios are equal. The
> rectangular pixels of the source are found to be the same in the
> project, as are the DAR and the number of pixels of WxH. In fact, the
> format of the project is the same as that of the source.
>
> source_4 (differentWxH; different SAR; same DAR) ==> non-square
> Pixels; flattened display; crop.
>           Because of WxH_source > WxH_project we will have a crop, even
> though the DAR is the same. This is because we go from DAR 1:1 to SAR
> 4:3 so the pixels are forced to extend horizontally. Since the DAR is
> the same there must be a crop of the source to fit within the smaller
> WxH of the project.
>
> So:
>
> same WxH ==> Correct display.
> WxH_source < WxH_project ==> flattened display; pad.
> WxH_source > WxH_project ==> flattened display; crop.
>
> SAR ==> determines the horizontal size of the pixels (only the
> horizontal base). It is the first of the factors that adapt the
> WxH_source to the WxH_project.
>
> DAR ==> determines the shape of the frame. It is the second of the
> factors that adapt the WxH_source to the WxH_project.
>
>  From Set Format you set the design (WxH and DAR) at this point the
> sources will have to adapt to this format, either by changing DAR or
> by changing the pixel shape via SAR.
>
>
> Do the results and explanations sound right to you?
>
> I leave the link if anyone wants to try the EDL and the four videos:
>
> https://www.dropbox.com/scl/fi/hvltphw3yq76mkquok1wq/Archivio.tar.gz?rlkey=bxko4ee6z0awr4epusnpyj8p5&dl=0
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cinelerra-gg.org/pipermail/cin/attachments/20240123/8f277305/attachment.htm>