[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: True Color Image Manipulation
The reason for using object graphics in this case are:
1) Rotations and scaling will be faster. It would be nice to line up an
object on the image with the cursor draging one color image. This part can
easily be done) direct colors. Now select the registered object as the
pivot for rotation. Then select a second object in one color and drag it
to the object's location in a second color (rotation and scale change).
Trying to update the image in direct graphics during the drag operation
would be much too slow.
2) We are actually working with more that three spectral bands. Additional
bands would be displayed simultaneously as color contour plots. Different
colors for different bands. In direct graphics it would be easy to add a
single black and white contour plot but not multiple color ones. We will
also want to be able to easily add other overlay's (e.g. rubber band boxes
and elliptical apertures). Also when scaling all bands simultaneously (e.g.
zooming), bookeeping is easier and with direct graphics we would have to
replot all the contours.
3) We will have at least two displays of the results updated
simulataneously. A full image and a zoomed image. Use of IDL objects
makes this easy. It would be much easier to use the existing IDL graphic
objects instead of constructing our own objects from scratch. In a sense,
if we used direct graphics we would be basically simulating the capabilities
of the IDL graphic objects (Meaning alot of extra coding) without the added
graphic speed available on the computer.
Maybe I can convince David Stern to fix the 2-D transformation for
IDLgrImage objects and allow the IDLgrImage object to contain a single
"David Fanning" <firstname.lastname@example.org> wrote in message
> Don J Lindler (email@example.com) writes:
> > Does anyone know a way to create a true color image using object
> > such that the three
> > color planes can be manipulated independently (translation, scaling,
> > rotation in 2 dimensions).
> > In other words, I need each color (red, green, and blue images) in
> > IDLgrModel objects.
> > I was able to use IDLgrImage objects with the channel set ('ff0000'x for
> > red, '00ff00'x for green,
> > and '0000ff',x for blue). These IDLgrImage objects were added to three
> > separate
> > model objects. I was able to translate and scale each model separately
> > not rotate them.
> > Rotation in 2 dimensions does not work properly for a IDLgrImage object.
> > Also this approach
> > has the problem that I need to supply all three color images to each
> > IDLgrImage object even
> > though I am only using one of them.
> > I contacted RSI support and they first suggested using an IDLgrPolygon
> > object with the Texture
> > maps set to the images. This allows rotation in all three dimensions
> > did not work. The channel
> > property of the IDLgrImage was ignored when used as a texture map and it
> > not solve the
> > problem of supplying three images to each IDLgrImage object.
> > I contacted RSI again and they suggested using a single IDLgrImage
> > and performing the
> > image rotations manually with the ROT function and reloading the images
> > the IDLgrImage
> > object with setproperty. This works but was 50% slower than just using
> > direct graphics instead
> > of object graphics.
> > I would appreciate any suggestions or solutions.
> I think you have pretty much covered the ground here.
> But tell me again, what compelling reason do you have
> to use object graphics in the first place?
> David Fanning, Ph.D.
> Fanning Software Consulting
> Phone: 970-221-0438 E-Mail: firstname.lastname@example.org
> Coyote's Guide to IDL Programming: http://www.dfanning.com/
> Toll-Free IDL Book Orders: 1-888-461-0155