Flesh out the video filter base class. Make it parse the input and output caps
and turn them into GstVideoInfo. Map buffers as video frames and pass them to
the transform functions.
This allows us to also implement the propose and decide_allocation vmethods.
Implement the transform size method as well.
Update subclasses with the new improvements.
With the new video bufferpool we can now implement the propose_allocation
vmethod on some video filter elements so that we can also use video metadata and
bufferpools when not operating in passthrough mode.
Adds a Lanczos-derived scaling method, which is rather slow, but very
high quality. Adds a few properties that can be used to tune various
scaling properties: sharpness, sharpen, envelope, dither. Not currently
Orcified, but was designed with that in mind.
Make a new GstVideoFormatinfo structure that contains the specific information
related to a format such as the number of planes, components, subsampling,
pixel stride etc. The result is that we are now able to introduce the concept of
components again in the API.
Use tables to specify the formats and its properties.
Use macros to get information about the video format description.
Move code to set strides, offsets and size into one function.
Remove methods that are not handled with the structures.
Add methods to retrieve pointers and strides to the components in the video.
Remove the GstVideoPlane structure and move the fields directly into the
GstVideoInfo structure. This makes things a little easier to read and also makes
it more likely that we can pass the stride array to external libraries.
If the second and next caps structures are a subset of the already existing
transformed caps we can safely skip them because we would transform them to
the same caps again.
When fixating caps, from_par should always be initialized
with a fixed value.
In case the fixation is from src to sink pad it was setting
the from par (srcpad par) to a fraction range, this patch initializes
it to 1/1, based on the assumption that missing PAR is 1/1.
https://bugzilla.gnome.org/show_bug.cgi?id=641952
Otherwise we're producing different caps and basetransform thinks that it
can't passthrough buffer allocations, etc.
In 0.11 all video caps really should have the PAR set...
This behaviour was not preferred and caused visible image quality
degradations. The real solution would be, to apply a real
deinterlacing filter before scaling the frames.
Fixes bug #615471.
Original commit message from CVS:
* gst/videoscale/gstvideoscale.c: (gst_video_scale_transform_caps):
Transform a given PAR to a range on the struct with the generic
height/width instead of the struct with the possibly restricted
height/width.
Original commit message from CVS:
* gst/videoscale/gstvideoscale.c: (gst_video_scale_transform_caps):
Prefer the given format if it contains something stricter than [1,MAX]
for height or width and only put a structure that requires rescaling
as second. This makes it possible to use videoscale in pipelines where
the source can actually produce the wanted height/width but usually
selects a different one from the requested.
Original commit message from CVS:
* gst/videoscale/gstvideoscale.c: (DEFAULT_PROP_METHOD):
Change default scaling method from nearest-neighbour to bilinear.
Original commit message from CVS:
* ext/vorbis/vorbisdec.c:
* gst/videoscale/gstvideoscale.c: (gst_video_scale_transform):
* sys/xvimage/xvimagesink.c: (gst_xvimagesink_show_frame):
Some debug and comment fixes.
* tests/examples/dynamic/addstream.c: (main):
Fix , to ;
Original commit message from CVS:
* gst/videoscale/gstvideoscale.c:
* gst/videoscale/vs_4tap.c: (vs_image_scale_4tap_Y):
* gst/videoscale/vs_image.c: (vs_image_scale_nearest_RGBA),
(vs_image_scale_linear_RGBA), (vs_image_scale_nearest_RGB),
(vs_image_scale_linear_RGB), (vs_image_scale_nearest_YUYV),
(vs_image_scale_linear_YUYV), (vs_image_scale_nearest_UYVY),
(vs_image_scale_linear_UYVY), (vs_image_scale_nearest_Y),
(vs_image_scale_linear_Y), (vs_image_scale_nearest_RGB565),
(vs_image_scale_linear_RGB565), (vs_image_scale_nearest_RGB555),
(vs_image_scale_linear_RGB555):
Support 1x1 images as input and output as for example the BBC HQ new
streams have 1x1 GIFs in the playlists for some reason.
Original commit message from CVS:
* gst/videoscale/gstvideoscale.c: (gst_video_scale_set_property),
(gst_video_scale_get_property), (gst_video_scale_transform_caps),
(gst_video_scale_transform):
Don't claim to be able to handle/transform caps that can't really
be handled by the currently selected scaling method (here: RGB or
packed YUV with 4-tap method). Also add locking to method property.
* tests/check/pipelines/simple-launch-lines.c: (setup_pipeline),
(test_basetransform_based):
Some test pipelines for the above (not entirely valgrind clean yet
apparently).
Original commit message from CVS:
* gst/videoscale/gstvideoscale.c:
A width and height of 1 makes us crash, so increase minimum size to
2x2 pixels until someone feels like fixing this (#404512).
Original commit message from CVS:
* gst/playback/gstplaybasebin.c: (check_queue_event):
Improve debug.
* gst/videoscale/gstvideoscale.c: (gst_video_scale_transform_caps):
Fix width and height range from 16 - 4096 to 1 - MAXINT, just like the
padtemplate caps. Refixes #357577.
Original commit message from CVS:
* gst/videoscale/Makefile.am:
* gst/videoscale/gstvideoscale.c:
* gst/videoscale/gstvideoscale.h:
* gst/videoscale/vs_4tap.c:
* gst/videoscale/vs_4tap.h:
* gst/videoscale/vs_image.c:
* gst/videoscale/vs_image.h:
* gst/videoscale/vs_scanline.c:
* gst/videoscale/vs_scanline.h:
Add a 4-tap image scaler. Theoretically looks much prettier.
The tap calculation could use some improvement.
Original commit message from CVS:
* gst/audioconvert/gstaudioconvert.c:
(gst_audio_convert_get_unit_size), (set_structure_widths):
Lower debug, use g_assert in _get_unit_size
* gst/audioresample/gstaudioresample.c:
(audioresample_get_unit_size):
* gst/ffmpegcolorspace/gstffmpegcolorspace.c:
(gst_ffmpegcsp_get_unit_size):
* gst/videoscale/gstvideoscale.c: (gst_video_scale_get_unit_size):
use g_assert in _get_unit_size
Original commit message from CVS:
* gst/audioresample/gstaudioresample.c: (audioresample_stop),
(audioresample_set_caps):
Don't leak references to the incoming caps. Clean them up when
stopping.
* gst/videoscale/gstvideoscale.c: (gst_video_scale_class_init),
(gst_video_scale_finalize):
Don't leak our temporary pixel buffer.
* tests/check/Makefile.am:
* tests/check/pipelines/simple-launch-lines.c: (run_pipeline),
(GST_START_TEST), (simple_launch_lines_suite):
Fix leaks and re-enable the test for valgrind checking.
Original commit message from CVS:
Patch by: Philip Jaegenstedt <philip at lysator dot liu dot se>
* gst/videoscale/gstvideoscale.c: (gst_video_scale_prepare_size),
(gst_video_scale_transform):
Make videoscale support RGBA, ARGB, BGRA and ABGR. Fixes#345131
Original commit message from CVS:
* gst/videoscale/gstvideoscale.c: (gst_video_scale_fixate_caps):
* sys/xvimage/xvimagesink.c: (gst_xvimagesink_setcaps):
Use the gstutil scaling function to preserve 64 bits while calculating
output width and height from the display-aspect-ratio. (A continuation
of #341542)
Original commit message from CVS:
* gst/videoscale/gstvideoscale.c: (gst_video_scale_fixate_caps):
* sys/xvimage/xvimagesink.c: (gst_xvimagesink_setcaps):
Fix the build.
Original commit message from CVS:
* gst-libs/gst/video/gstvideofilter.c: (gst_video_filter_init):
* gst-libs/gst/video/gstvideosink.c: (gst_video_sink_init):
* gst/ffmpegcolorspace/gstffmpegcolorspace.c: (gst_ffmpegcsp_init):
* gst/videoscale/gstvideoscale.c: (gst_video_scale_class_init),
(gst_video_scale_init), (gst_video_scale_src_event):
Re-enable QoS after the release.
Rework videoscale to use the base class src_event handler.
Original commit message from CVS:
2005-11-18 Julien MOUTTE <julien@moutte.net>
* gst/videoscale/gstvideoscale.c: (gst_videoscale_class_init),
(gst_videoscale_fixate_caps): Introduce back caps fixate with
handling of PAR.
Original commit message from CVS:
* check/Makefile.am:
* check/pipelines/simple_launch_lines.c: (setup_pipeline),
(run_pipeline), (GST_START_TEST), (simple_launch_lines_suite):
Add extra tests for basetransform based components.
Comment out the test_element_negotiation test until we decide
if it's testing correct behaviour.
* ext/libvisual/visual.c: (gst_visual_init), (get_buffer),
(gst_visual_chain), (gst_visual_change_state):
Slightly more correct but still bogus timestamping.
Fix state change function.
* gst/audioconvert/gstaudioconvert.c:
(gst_audio_convert_class_init):
* gst/audioresample/gstaudioresample.c:
* gst/ffmpegcolorspace/gstffmpegcolorspace.c:
(gst_ffmpegcsp_class_init):
* gst/videoscale/gstvideoscale.c: (gst_videoscale_class_init),
(gst_videoscale_prepare_size), (gst_videoscale_set_caps),
(gst_videoscale_prepare_image):
* gst/volume/gstvolume.c: (gst_volume_class_init),
(volume_transform_ip):
Basetransform updates. Enable passthrough modes.
* sys/ximage/ximagesink.c: (gst_ximage_buffer_init),
(gst_ximagesink_renegotiate_size), (gst_ximagesink_xcontext_get),
(gst_ximagesink_setcaps), (gst_ximagesink_buffer_alloc):
Negotiation fix that allows the window to return to the original
size and renegotiate passthrough upstream. Extra debug output.
Original commit message from CVS:
* gst/videoscale/gstvideoscale.c: (gst_videoscale_init),
(gst_videoscale_prepare_size), (parse_caps),
(gst_videoscale_set_caps), (gst_videoscale_get_size),
(gst_videoscale_prepare_image), (gst_videoscale_transform_ip),
(gst_videoscale_transform):
* gst/videoscale/gstvideoscale.h:
Refactor, make use of BaseTranform really well.
