The demosaic code first expands the buffer datatype to uint16, and then
shifts the data left so that the 8, 10 and 12 bitspp formats all become
16 bitspp.
It then, eventually, uses np.einsum to calculate averages, but this
averaging sums multiple uint16 values together, and stores them in
uint16 storage. As in the first step we shifted the values left,
possibly getting values close to the maximum of uint16 range, we, of
course, overflow when summing them together. This leads to rather bad
looking images.
Fix this by dropping the original shift. It serves no purpose, and is
probably a remnant of some early testing code. This way the largest
numbers we are summing together are 12 bit values, and as we use a 3x3
window from which we fetch values, for a single rgb plane, the max
number of 12 bit values is 5 (for green). Sum of 5 12 bit values is well
below the 16 bit maximum.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Currently, if there are multiple streams, the --metadata flag will print
the metadata for each request multiple times.
Moving the metadata print logic outside the stream for loop this will no
longer occur.
Signed-off-by: Daniel Oakley <daniel.oakley@ideasonboard.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The python bindings are missing the ability to read the sequence number
of the Request object from the public API.
Expose the objects sequence number on the pybind11 surfaces to support
applications reading this value.
Signed-off-by: Matthew Goodman <matt@exclosure.io>
[Kieran: Revised commit message]
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
The Python bindings will now accept, or return, a list or tuple of
libcamera.Rectangle objects for such controls.
This had previously been omitted, but now we have, for example, the
AfWindows control which requires this feature.
Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Such controls can now be created when a control doesn't have a
reasonable or obvious default value. We support them using Python's
"None" value, rather than generating a runtime error.
Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Christian Rauch <Rauch.Christian@gmx.de>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Miscellaneous files that are not considered copyrightable are missing an
SPDX-License-Identifier header. Fix this by adding a header listing the
CC0-1.0 license.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
The JPEG color space is badly name, as the JPEG specification (ITU-T
T.81) doesn't define any particular color space:
The interchange format does not specify a complete coded image
representation. Application-dependent information, e.g. colour
space, is outside the scope of this Specification.
The JFIF specification (ITU-T T.871) is clearer as it requires ITU-R
BT.601 YCbCr encoding and a full quantization range:
The interpretations of Y, CB, and CR are derived from the E'Y, E'Cb,
and E'Cr signals defined in the 625-line specification of Rec. ITU-R
BT.601, but these signals are normalized so as to permit the usage of
the full range of 256 levels of the 8-bit binary encoding of the Y
component.
It however doesn't specify color primaries or a transfer function
explicitly. It only mentions the latter when describing the conversion
from YCbCr to RGB:
The inverse relationship for computing full scale 8-bit per colour
channel gamma pre-corrected RGB values (following Rec. ITU-R BT.601
gamma pre-correction and colour primary specifications) from YCbCr
colours (with 256 levels per component) can be computed as follows:
[...]
Given that ITU-R BT.601-5 (1995) didn't specify color primaries or a
transfer function, and that the later ITU-R BT.601-7 (2011) version
specifies color primaries for the 625-line variant that do not match
sRGB, the JPEG color space in libcamera is badly named. This is
confirmed by ITU-T T.871:
As this Recommendation | International Standard is based on the prior
informally-circulated JFIF version 1.02 specification that was
produced in 1992, which referenced Rec. ITU-R BT.601 (formerly CCIR
601), it references that specification for definition of the E'Y,
E'Cb, and E'Cr signals that correspond to the YCBCR values specified
herein. However, since the development of the prior JFIF version 1.02
specification, additional industry specifications have been developed,
Rec. ITU-R BT.601 has been updated, and common industry practice has
emerged which often follows the sYCC specification in IEC
61966-2-1/Amd.1. The difference between the use of the colour
interpretation specification in this Recommendation | International
Standard and that of the sYCC specification may be considered
negligible in practice.
Rename the color space to sYCC, as its definition matches the sYCC
standard, and indicate that it is typically used to encode JPEG images.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Umang Jain <umang.jain@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Blocking wait can be easily implemented on top in Python, so rather than
supporting only blocking reads, or supporting both non-blocking and
blocking reads, let's support only non-blocking reads.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Use libcamera's Mutex and MutexLocker instead of the std versions to get
thread safety annotations.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Set EFD_CLOEXEC on eventfd to avoid fd leaking.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Use UniqueFD to automate the eventfd lifetime management.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Wrap the CameraManager with a PyCameraManager class and move the related
code inside the new class.
This helps understanding the life times of the used-to-be global
variables, gets rid of static handleRequestCompleted function, and
allows us to simplify the binding code as the more complex pieces are
inside the class.
There should be no user visible functional changes.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
We define -DLIBCAMERA_BASE_PRIVATE to get access to libcamera private
headers, but the correct way to do this is to have a meson dependency to
libcamera_private.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Clean up the py_main.cpp a bit by moving the ControlValue helpers to a
separate file.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Add Python logging category, and use it in handleRequestCompleted().
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
-C flag is supposed to affect only the camera that was previously
defined in the arguments. That's not the case, and, e.g.:
cam.py -c2 -C -c3
causes camera 3 to start capturing, but it stops after the initial
Requests have been completed.
Fix the issue by filtering out camera contexts that do not have -C
defined.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Add a Python version of simple-cam from:
https://git.libcamera.org/libcamera/simple-cam.git
Let's keep this in the libcamera repository until the Python API has
stabilized a bit more, and then we could move this to the simple-cam
repo.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Add a slightly more complex, and I think a more realistic, example,
where the script reacts to events and re-queues the buffers.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Add an example to showcase the more-or-less minimal capture case.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
We always call CameraManager.read_event() and
CameraManager.get_ready_requests(), so to simplify the use merge the
read_event() into the get_ready_requests().
This has the side effect that get_ready_requests() will now block if
there is no event ready. If we ever need to call get_ready_requests() in
a polling manner we will need a new function which behaves differently.
However, afaics the only sensible way to manage the event loop is to use
select/poll on the eventfd and then call get_ready_requests() once,
which is the use case what the current merged function supports.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Instead of doing an mmap and munmap every time a Request is complete,
mmap all the buffers once at the start of the program.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Implement non-contextmanager use to MappedFrameBuffer so that we can
either:
with MappedFrameBuffer(fb) as mfb:
...
or
mfb = MappedFrameBuffer(fb)
mfb.mmap()
...
mfb.munmap()
While at it, improve the error handling a bit.
Note that the mmap() returns self. In other words, one can do this:
mfb = MappedFrameBuffer(fb).mmap()
...
mfb.munmap()
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Implement FrameBufferPlane class and adjust the methods and uses
accordingly.
Note that we don't expose the fd as a SharedFD, but as an int.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Add FrameMetadataPlane class and adjust the methods and uses
accordingly.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
We can use Qt directly to accomplish the same as we do with PIL.
A minor downside is that loading MJPEG frame with Qt produces a "Corrupt
JPEG data" warning. The resulting picture looks fine, though. So add a
message handler to ignore that warning.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Move conversion functions from cam_qt.py to helpers.py to clean up the
code and so that they can be used from other cam renderers.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
We create symlinks from the Python bindings build dir to the source dir
so that 1) the build dir can be used to use the bindings, and 2) to
allow modifications of the source .py files to be used right away
without rebuilding.
The symlinks were recently fixed and changed to use absolute paths.
However, absolute paths ruin one main use case I have: using the
bindings from the build dir via nfs from an ARM device.
So move back to relative paths, but accomplish this with the --relative
parameter for ln, instead of guessing the right relative path as was
done before the above-mentioned fix.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Generate bindings for properties in a very similar way as done for
controls. We do need to distinguish between the two, and thus I added
--properties flag to gen-py-controls.py.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Perhaps it's better to have a more descriptive name here. I also
considered just renaming 'efd' to 'fd', but 'event_fd' won.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Add ControlInfo class and change the controls related methods to
resemble the C++ API (e.g. no more string based control methods).
We don't implement ControlList or ControlInfoMap but just expose the
same data via standard Python dict.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The Python bindings controls generation was not very good. It only
covered the enums and they were in the main namespace.
This adds the controls somewhat similarly to the C++ side. We will have
e.g.:
libcamera.controls.Brightness
libcamera.controls.AeMeteringModeEnum.CentreWeighted
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Convert ctx and state dicts to classes. No functional changes.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Drop WA_ShowWithoutActivating from the Qt renderers. I added the flag
during development phase as I didn't want the window to take the focus,
but it should be removed now.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Add a few type hints and (minimal) docs to MappedFrameBuffer.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Move MappedFrameBuffer to libcamera.utils, instead of extending
FrameBuffer class with a new mmap() method. This keeps us more aligned
to the C++ API.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Add CameraManager.read_event() so that the user does not need to call
os.read().
We use eventfd, and we must always read 8 bytes. Hiding that inside
read_event() makes sense.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
"None" is not a valid name for an enum value, so change it to "Null".
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Stripping 'SceneFlicker' prefix from the enum value names leads to
'50Hz' and '60Hz', which are not valid names. So add another heuristics
to keep the prefix for SceneFlicker.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Use the new libcamera.formats in cam_qt.py.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Generate a list of pixel formats under 'libcamera.formats'.
The 'formats' is a "dummy" container class, the only purpose of which is
to contain the read-only pixel format properties.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Try to be more consistent with the names, and include "control" in all
the names.
Also drop a useless "using namespace libcamera" and only include
"control_ids.h".
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Having the underscore makes the names more readable, especially when
there are multiple words in the name.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Now that we have proper geometry classes in the Python bindings, change
the existing bindings and the .py files accordingly.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Add libcamera's geometry classes to the Python bindings.
Note that this commit only adds the classes, but they are not used
anywhere yet.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Implement PixelFormat bindings properly with a PixelFormat class. Change
the bindings to use the new class instead of a string.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
