Jonathan Thomas
b094ac8469
Fixing issue with incorrect image size caches, which results in blurry/smudgy scaling on certain things.
2017-07-27 02:25:20 -05:00
Jonathan Thomas
1743558f99
Improved image caching logic, to better estimate max image sizes possible, based on clip scale and keyframe settings... so we are always dealing with the smallest possible frame sizes for performance (without losing quality)
2016-09-17 17:14:27 -05:00
Jonathan Thomas
183b0714e2
Added a new optimized constructor to some readers (FFmpegReader, QtImageReader, ImageReader) to not Open() in the constructor, which dramatically speeds up creating hundreds of Readers (i.e. when opening a project). This is really only useful when inflating the reader with Json right after you instantiate it...
2016-09-16 17:43:26 -05:00
Jonathan Thomas
7f347eb1ca
Added SetMaxSize for image optimizations in QImageReader and FFmpegReader, which lets the timeline pass down the max size to all clips and readers, so they can optionally optimize the size of images (especially useful for optimizing preview performance). Removed convoluted image scaling code in FFmpegReader, and replaced with simpler version. Also, fixed a few regressions from the new Caching code, primarily a crash when reaching the end of the last clip on the timeline.
2016-09-14 04:11:12 -05:00
Jonathan Thomas
66c5ced902
Windows now passes most unit tests. Added header macro to assist on passing correct test media paths to unit tests. Also added a new info parameter has_single_image, which is used to optomize the frame mapper class for images with long durations.
2015-09-28 22:05:50 -05:00
Jonathan Thomas
bfa050409c
Improved support for missing frames and invalid PTS (which result in gaps and duplicate frames). Also changed frame number to larger data type, and did some code clean-up and refactoring. Less crashes, and more video support!
2015-08-24 01:05:48 -05:00
Jonathan Thomas
b612f3339d
Replaced ImagMagick with QImage on almost all key methods and classes. Reprogrammed all effects and the entire rendering pipeline to use QImage and QTransforms, primarily for increases in speed and stability. libopenshot is more than 10X faster on many of the most CPU heavy tasks. This was a huge change, and still has a few minor issues relating to BlackMagick Decklink and Text rendering.... which should be resolved very soon.
...
Also, much work has been done on memory management / leak detection, and optimizations with multi-threading... including a new thread cacher class used by the video playback (which is smoother than ever).
2015-06-01 00:20:14 -07:00
