diff options
| author | Carl Hetherington <cth@carlh.net> | 2021-10-03 21:22:14 +0200 |
|---|---|---|
| committer | Carl Hetherington <cth@carlh.net> | 2021-10-03 23:41:02 +0200 |
| commit | 969906f2dd6c5c144781861f53e2a0f6baefb9a3 (patch) | |
| tree | d4091610b2c0582f52a8d484fd7a36a9c4878111 /src/lib/util.cc | |
| parent | 70b72b53eab0f247eb4dc605a2d669d4adb4e469 (diff) | |
Don't scale chroma subsampled images to sizes that don't align with the subsampling (#1872).v2.15.166
There's a slightly odd effect when scaling e.g. YVU420 images to
odd and then even widths - there's a small but visible luminance
shift. I don't know why this happens, but keeping the scaling
sizes locked to the subsampling seems to help.
Diffstat (limited to 'src/lib/util.cc')
| -rw-r--r-- | src/lib/util.cc | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/src/lib/util.cc b/src/lib/util.cc index d3af74376..78ed8da99 100644 --- a/src/lib/util.cc +++ b/src/lib/util.cc @@ -1111,7 +1111,7 @@ linear_to_db (double linear) dcp::Size -scale_for_display (dcp::Size s, dcp::Size display_container, dcp::Size film_container) +scale_for_display (dcp::Size s, dcp::Size display_container, dcp::Size film_container, PixelQuanta quanta) { /* Now scale it down if the display container is smaller than the film container */ if (display_container != film_container) { @@ -1122,6 +1122,7 @@ scale_for_display (dcp::Size s, dcp::Size display_container, dcp::Size film_cont s.width = lrintf (s.width * scale); s.height = lrintf (s.height * scale); + s = quanta.round (s); } return s; |