dcpomatic/src/lib/image.cc, branch master DCP-o-matic DCP tools https://git.carlh.net/cgit/dcpomatic/atom?h=master 2021-05-06T19:59:56Z Add pixel format 0 (AV_PIX_FMT_YUV420P) to make_part_black(). 2021-05-06T19:59:56Z Carl Hetherington cth@carlh.net 2021-05-02T23:38:11Z urn:sha1:52f6790026040e17237ba8d7865cda3e0e239b62 Remainder of fix for #1984. Backported-from-commit: 0aabe4060ea4bad7c7caac633aef0737fccff8c2 Backported-from-branch: 2.15.x Add pixel format 66 (AV_PIX_FMT_YUV422P10LE) to make_part_black(). 2021-05-06T19:49:41Z Carl Hetherington cth@carlh.net 2021-05-02T09:30:16Z urn:sha1:29ad9ebaba113f0300b0c94fd1fa29f67e9f24a6 Part of fix for #1984. Backported-from-commit: 2aa6fd88e6d334c040d421938e425bd2f89983a7 Backported-from-branch: 2.15.x Fix corrupted image when over-cropping black filler frames. 2020-11-24T23:11:55Z Carl Hetherington cth@carlh.net 2020-11-24T23:11:55Z urn:sha1:accdec63a79b43c6349597b15243dc41e521521d FFmpegDecoder can emit small black frames (128x128 pixels) when it wants to fill in a gap. Image::crop_scale_window would do the wrong thing if we then applied a crop of greater than 128 in either direction; though cropped_size is correctly clamped, the crop value itself was not and is used to calculate the input data pointers. This would result in random frames, usually at the end of DCPs, often made up of blurry colour washes. Hack to fix image luminance when cropping subsampled images (#1872). 2020-11-19T21:08:21Z Carl Hetherington cth@carlh.net 2020-11-19T21:07:39Z urn:sha1:2456409a8db781d9ef505fc9e5f9290276ce6807 Without this the luminance varies slightly as you crop by odd and even amounts (for YUV420 images). Fix cropping of subsampled images. 2020-11-17T08:59:30Z Carl Hetherington cth@carlh.net 2020-11-16T23:00:50Z urn:sha1:27c0e43d75218b926068fff3d6d1364b476e56ff The calculations for how to crop subsampled components of YUV images were wrong, causing strange effects like misregistration of colour components in cropped images. Should fix #1872. Add some lrintf() calls to be sure. 2020-11-16T23:18:02Z Carl Hetherington cth@carlh.net 2020-11-16T22:58:53Z urn:sha1:451e5465bbd10827dafe76b35704102e589c0980 Use 64-byte alignment for aligned image memory. 2020-11-16T23:17:52Z Carl Hetherington cth@carlh.net 2020-11-16T22:56:11Z urn:sha1:cdf9ebe4a827d8d2d6d89e34bad0c976b04834b7 FFmpeg uses this values if AVX512 is available, and with only 32-byte alignment I am seeing strange scaling effects whereby crop_scale_window_test7 gives black bars down the right side of cropped images (when run on an i7 7700). Fix incorrect images when cropping without stretch. 2019-11-16T21:46:39Z Carl Hetherington cth@carlh.net 2019-11-04T15:38:14Z urn:sha1:ed4fc06db6957b2b63b2400a737f47c18a1003be Always overallocate images so that Image::crop_scale_window is always safe from over-reading buffers. Relates to #1654 and probably #1653. Backported from 7b0372776ac4da6a8e4ff29f41a4f08b9b4de506 in v2.15.x. Fix "cannot handle pixel format 26 during make_transparent()" (part of #1545). 2019-05-03T14:00:41Z Carl Hetherington cth@carlh.net 2019-05-03T14:00:19Z urn:sha1:46df210e1c25c3cdae664390efac8e60714ad635 Fix or remove several broken pixel formats in Image::fade and add 2019-04-25T15:19:25Z Carl Hetherington cth@carlh.net 2019-04-25T15:19:25Z urn:sha1:11efa93e15b694b8ea6f0a2bc68c87503cc570bb unit tests for the remainder. Fixes #1532.