linux/Documentation/DocBook/media/v4l/selection-api.xml

<section id="selection-api">

  <title>Experimental API for cropping, composing and scaling</title>

      <note>
	<title>Experimental</title>

	<para>This is an <link linkend="experimental">experimental</link>
interface and may change in the future.</para>
      </note>

  <section>
    <title>Introduction</title>

<para>Some video capture devices can sample a subsection of a picture and
shrink or enlarge it to an image of arbitrary size. Next, the devices can
insert the image into larger one. Some video output devices can crop part of an
input image, scale it up or down and insert it at an arbitrary scan line and
horizontal offset into a video signal. We call these abilities cropping,
scaling and composing.</para>

<para>On a video <emphasis>capture</emphasis> device the source is a video
signal, and the cropping target determine the area actually sampled. The sink
is an image stored in a memory buffer.  The composing area specifies which part
of the buffer is actually written to by the hardware. </para>

<para>On a video <emphasis>output</emphasis> device the source is an image in a
memory buffer, and the cropping target is a part of an image to be shown on a
display. The sink is the display or the graphics screen. The application may
select the part of display where the image should be displayed. The size and
position of such a window is controlled by the compose target.</para>

<para>Rectangles for all cropping and composing targets are defined even if the
device does supports neither cropping nor composing. Their size and position
will be fixed in such a case. If the device does not support scaling then the
cropping and composing rectangles have the same size.</para>

  </section>

    <section>
      <title>Selection targets</title>

      <para>
      <figure id="sel-targets-capture">
	<title>Cropping and composing targets</title>
	<mediaobject>
	  <imageobject>
	    <imagedata fileref="selection.png" format="PNG" />
	  </imageobject>
	  <textobject>
	    <phrase>Targets used by a cropping, composing and scaling
            process</phrase>
	  </textobject>
	</mediaobject>
      </figure>
      </para>

      <para>See <xref linkend="v4l2-selection-targets" /> for more
    information.</para>
    </section>

  <section>

  <title>Configuration</title>

<para>Applications can use the <link linkend="vidioc-g-selection">selection
API</link> to select an area in a video signal or a buffer, and to query for
default settings and hardware limits.</para>

<para>Video hardware can have various cropping, composing and scaling
limitations. It may only scale up or down, support only discrete scaling
factors, or have different scaling abilities in the horizontal and vertical
directions. Also it may not support scaling at all. At the same time the
cropping/composing rectangles may have to be aligned, and both the source and
the sink may have arbitrary upper and lower size limits. Therefore, as usual,
drivers are expected to adjust the requested parameters and return the actual
values selected. An application can control the rounding behaviour using <link
linkend="v4l2-selection-flags"> constraint flags </link>.</para>

   <section>

   <title>Configuration of video capture</title>

<para>See figure <xref linkend="sel-targets-capture" /> for examples of the
selection targets available for a video capture device.  It is recommended to
configure the cropping targets before to the composing targets.</para>

<para>The range of coordinates of the top left corner, width and height of
areas that can be sampled is given by the <constant>V4L2_SEL_TGT_CROP_BOUNDS</constant>
target. It is recommended for the driver developers to put the
top/left corner at position <constant>(0,0)</constant>.  The rectangle's
coordinates are expressed in pixels.</para>

<para>The top left corner, width and height of the source rectangle, that is
the area actually sampled, is given by the <constant>V4L2_SEL_TGT_CROP</constant>
target. It uses the same coordinate system as <constant>V4L2_SEL_TGT_CROP_BOUNDS</constant>.
The active cropping area must lie completely inside the capture boundaries. The
driver may further adjust the requested size and/or position according to hardware
limitations.</para>

<para>Each capture device has a default source rectangle, given by the
<constant>V4L2_SEL_TGT_CROP_DEFAULT</constant> target. This rectangle shall
over what the driver writer considers the complete picture.  Drivers shall set
the active crop rectangle to the default when the driver is first loaded, but
not later.</para>

<para>The composing targets refer to a memory buffer. The limits of composing
coordinates are obtained using <constant>V4L2_SEL_TGT_COMPOSE_BOUNDS</constant>.
All coordinates are expressed in pixels. The rectangle's top/left
corner must be located at position <constant>(0,0)</constant>. The width and
height are equal to the image size set by <constant>VIDIOC_S_FMT</constant>.
</para>

<para>The part of a buffer into which the image is inserted by the hardware is
controlled by the <constant>V4L2_SEL_TGT_COMPOSE</constant> target.
The rectangle's coordinates are also expressed in the same coordinate system as
the bounds rectangle. The composing rectangle must lie completely inside bounds
rectangle. The driver must adjust the composing rectangle to fit to the
bounding limits. Moreover, the driver can perform other adjustments according
to hardware limitations. The application can control rounding behaviour using
<link linkend="v4l2-selection-flags"> constraint flags</link>.</para>

<para>For capture devices the default composing rectangle is queried using
<constant>V4L2_SEL_TGT_COMPOSE_DEFAULT</constant>. It is usually equal to the
bounding rectangle.</para>

<para>The part of a buffer that is modified by the hardware is given by
<constant>V4L2_SEL_TGT_COMPOSE_PADDED</constant>. It contains all pixels
defined using <constant>V4L2_SEL_TGT_COMPOSE</constant> plus all
padding data modified by hardware during insertion process. All pixels outside
this rectangle <emphasis>must not</emphasis> be changed by the hardware. The
content of pixels that lie inside the padded area but outside active area is
undefined. The application can use the padded and active rectangles to detect
where the rubbish pixels are located and remove them if needed.</para>

   </section>

   <section>

   <title>Configuration of video output</title>

<para>For output devices targets and ioctls are used similarly to the video
capture case. The <emphasis>composing</emphasis> rectangle refers to the
insertion of an image into a video signal. The cropping rectangles refer to a
memory buffer. It is recommended to configure the composing targets before to
the cropping targets.</para>

<para>The cropping targets refer to the memory buffer that contains an image to
be inserted into a video signal or graphical screen. The limits of cropping
coordinates are obtained using <constant>V4L2_SEL_TGT_CROP_BOUNDS</constant>.
All coordinates are expressed in pixels. The top/left corner is always point
<constant>(0,0)</constant>.  The width and height is equal to the image size
specified using <constant>VIDIOC_S_FMT</constant> ioctl.</para>

<para>The top left corner, width and height of the source rectangle, that is
the area from which image date are processed by the hardware, is given by the
<constant>V4L2_SEL_TGT_CROP</constant>. Its coordinates are expressed
in in the same coordinate system as the bounds rectangle. The active cropping
area must lie completely inside the crop boundaries and the driver may further
adjust the requested size and/or position according to hardware
limitations.</para>

<para>For output devices the default cropping rectangle is queried using
<constant>V4L2_SEL_TGT_CROP_DEFAULT</constant>. It is usually equal to the
bounding rectangle.</para>

<para>The part of a video signal or graphics display where the image is
inserted by the hardware is controlled by <constant>V4L2_SEL_TGT_COMPOSE</constant>
target.  The rectangle's coordinates are expressed in pixels. The composing
rectangle must lie completely inside the bounds rectangle.  The driver must
adjust the area to fit to the bounding limits.  Moreover, the driver can
perform other adjustments according to hardware limitations.</para>

<para>The device has a default composing rectangle, given by the
<constant>V4L2_SEL_TGT_COMPOSE_DEFAULT</constant> target. This rectangle shall cover what
the driver writer considers the complete picture. It is recommended for the
driver developers to put the top/left corner at position <constant>(0,0)</constant>.
Drivers shall set the active composing rectangle to the default
one when the driver is first loaded.</para>

<para>The devices may introduce additional content to video signal other than
an image from memory buffers.  It includes borders around an image. However,
such a padded area is driver-dependent feature not covered by this document.
Driver developers are encouraged to keep padded rectangle equal to active one.
The padded target is accessed by the <constant>V4L2_SEL_TGT_COMPOSE_PADDED</constant>
identifier.  It must contain all pixels from the <constant>V4L2_SEL_TGT_COMPOSE</constant>
target.</para>

   </section>

   <section>

     <title>Scaling control</title>

<para>An application can detect if scaling is performed by comparing the width
and the height of rectangles obtained using <constant>V4L2_SEL_TGT_CROP</constant>
and <constant>V4L2_SEL_TGT_COMPOSE</constant> targets. If
these are not equal then the scaling is applied. The application can compute
the scaling ratios using these values.</para>

   </section>

  </section>

  <section>

    <title>Comparison with old cropping API</title>

<para>The selection API was introduced to cope with deficiencies of previous
<link linkend="crop"> API</link>, that was designed to control simple capture
devices. Later the cropping API was adopted by video output drivers. The ioctls
are used to select a part of the display were the video signal is inserted. It
should be considered as an API abuse because the described operation is
actually the composing.  The selection API makes a clear distinction between
composing and cropping operations by setting the appropriate targets.  The V4L2
API lacks any support for composing to and cropping from an image inside a
memory buffer.  The application could configure a capture device to fill only a
part of an image by abusing V4L2 API.  Cropping a smaller image from a larger
one is achieved by setting the field
&v4l2-pix-format;<structfield>::bytesperline</structfield>.  Introducing an image offsets
could be done by modifying field &v4l2-buffer;<structfield>::m_userptr</structfield>
before calling <constant>VIDIOC_QBUF</constant>. Those
operations should be avoided because they are not portable (endianness), and do
not work for macroblock and Bayer formats and mmap buffers.  The selection API
deals with configuration of buffer cropping/composing in a clear, intuitive and
portable way.  Next, with the selection API the concepts of the padded target
and constraints flags are introduced.  Finally, &v4l2-crop; and &v4l2-cropcap;
have no reserved fields. Therefore there is no way to extend their functionality.
The new &v4l2-selection; provides a lot of place for future
extensions.  Driver developers are encouraged to implement only selection API.
The former cropping API would be simulated using the new one.</para>

  </section>

   <section>
      <title>Examples</title>
      <example>
	<title>Resetting the cropping parameters</title>

	<para>(A video capture device is assumed; change
<constant>V4L2_BUF_TYPE_VIDEO_CAPTURE</constant> for other devices; change target to
<constant>V4L2_SEL_TGT_COMPOSE_*</constant> family to configure composing
area)</para>

	<programlisting>

	&v4l2-selection; sel = {
		.type = V4L2_BUF_TYPE_VIDEO_CAPTURE,
		.target = V4L2_SEL_TGT_CROP_DEFAULT,
	};
	ret = ioctl(fd, &VIDIOC-G-SELECTION;, &amp;sel);
	if (ret)
		exit(-1);
	sel.target = V4L2_SEL_TGT_CROP;
	ret = ioctl(fd, &VIDIOC-S-SELECTION;, &amp;sel);
	if (ret)
		exit(-1);

        </programlisting>
      </example>

      <example>
	<title>Simple downscaling</title>
	<para>Setting a composing area on output of size of <emphasis> at most
</emphasis> half of limit placed at a center of a display.</para>
	<programlisting>

	&v4l2-selection; sel = {
		.type = V4L2_BUF_TYPE_VIDEO_OUTPUT,
		.target = V4L2_SEL_TGT_COMPOSE_BOUNDS,
	};
	struct v4l2_rect r;

	ret = ioctl(fd, &VIDIOC-G-SELECTION;, &amp;sel);
	if (ret)
		exit(-1);
	/* setting smaller compose rectangle */
	r.width = sel.r.width / 2;
	r.height = sel.r.height / 2;
	r.left = sel.r.width / 4;
	r.top = sel.r.height / 4;
	sel.r = r;
	sel.target = V4L2_SEL_TGT_COMPOSE;
	sel.flags = V4L2_SEL_FLAG_LE;
	ret = ioctl(fd, &VIDIOC-S-SELECTION;, &amp;sel);
	if (ret)
		exit(-1);

        </programlisting>
      </example>

      <example>
	<title>Querying for scaling factors</title>
	<para>A video output device is assumed; change
<constant>V4L2_BUF_TYPE_VIDEO_OUTPUT</constant> for other devices</para>
	<programlisting>

	&v4l2-selection; compose = {
		.type = V4L2_BUF_TYPE_VIDEO_OUTPUT,
		.target = V4L2_SEL_TGT_COMPOSE,
	};
	&v4l2-selection; crop = {
		.type = V4L2_BUF_TYPE_VIDEO_OUTPUT,
		.target = V4L2_SEL_TGT_CROP,
	};
	double hscale, vscale;

	ret = ioctl(fd, &VIDIOC-G-SELECTION;, &amp;compose);
	if (ret)
		exit(-1);
	ret = ioctl(fd, &VIDIOC-G-SELECTION;, &amp;crop);
	if (ret)
		exit(-1);

	/* computing scaling factors */
	hscale = (double)compose.r.width / crop.r.width;
	vscale = (double)compose.r.height / crop.r.height;

	</programlisting>
      </example>

   </section>

</section>