Oh, interesting problem!
The resize() description says:
/// Set `dst`, over the region of interest, to be a resized version of the
/// corresponding portion of `src` (mapping such that the "full" image
/// window of each correspond to each other, regardless of resolution).
/// If `dst` is not yet initialized, it will be sized according to `roi`.
bool OIIO_API resize (ImageBuf &dst, const ImageBuf &src,
string_view filtername = "", float filterwidth=0.0f,
ROI roi={}, int nthreads=0);
It certainly doesn't make it easy to do what you want -- I think this points
out the need to have a variety of this call that takes a srcroi and a dstroi
and resizes between them, instead of the rescaling factor assuming that the two
images correspond to each other according to their "full" windows (in OIIO
lingo; this is what OpenEXR calls the "display window").
But maybe we can work around this creatively.
The first approach I can think of might work like this:
1. Save the full widows of both images with ImageBuf::roi_full().
2. Reset their full windows (using ImageBuf::set_roi_full()) of the src to its
data window (ImageBuf::roi()), and the dst to the subregion roi where you want
it to end up. This is not expensive, it's just metadata, doesn't copy any
pixels.
3. Now do the resize: ImageBufAlgo::resize(dst, src, filtername, filterwidth,
subset_roi); Only the dst pixels in subset_roi will be altered, but because we
carefully set the full window sizes, it ought to get a correct copy of the
whole src image.
4. Now restore the full windows to their original values (saved in step 1).
There should definitely be a way to do this all in one step, sorry that there
is not.
Another way to do this might be to use ImageBufAlgo::warp(), which works the
same way as resize -- it assumes that the "full" windows of the two images
correspond to each other. But you give it a 3x3 matrix (a 2D homogeneous
transformation), and with that you could arbitrarily map which area of src
corresponds to that subregion of dst that you are writing to. You'd probably
need to sit down with pencil and paper to figure out how to construct just the
right matrix, but it should work.
> On Sep 28, 2020, at 7:00 AM, Simon Smith <[email protected]> wrote:
>
> Hi,
>
> I’m trying to work out if/how I can resize one ImageBuf into a sub-region of
> another without any intermediate ImageBuf's to hold the resize.
> Essentially I want to take an image and create a new one of the same pixel
> size, but with a border around it.
>
> I’m doing this in memory with C++.
> Does anyone have any ideas on how this could be done without the need for
> extra memory/copy?
> I can’t immediately see anything in the API - they all seem to require a
> smaller dst ImageBuf for the result of the resize - but I’m sure there must
> be one in there somewhere!
>
>
> Best Regards,
> Simon
>
> ---
> Simon C Smith
> Co-Founder & CTO
> www.lightmap.co.uk
>
> Email confidentiality notice:
> This message is private and confidential. If you have received this message
> in error, please notify us and remove it from your system.
>
> _______________________________________________
> Oiio-dev mailing list
> [email protected]
> http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org
--
Larry Gritz
[email protected]
_______________________________________________
Oiio-dev mailing list
[email protected]
http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org
