>Psychtoolbox>PsychOpenGL

PsychOpenGL – OpenGL support for Matlab.

Psychtoolbox allows you to directly call low-level OpenGL commands from
the Matlab environment in nearly the same way as you could do from native
C or C++ code. This way you can code and use image manipulation
algorithms and 3D graphics algorithms in Matlab, utilizing the power of
modern graphics hardware by calling OpenGL functions.

Access to OpenGL from Matlab is provided by the “Matlab OpenGL toolbox”
(MOGL), whose original OS/X version was developed, implemented and
contributed to Psychtoolbox under the MIT license by Prof. Richard F.
Murray, University of York, Canada. (The code was under GPL license until
2010, but has been relicensed to the more permissive MIT license in 2011).

MOGL provides one Matlab wrapper M-File for each corresponding OpenGL
function. The wrapper file calls into a special MEX file (moglcore) which
implements the C-language interface to OpenGL. The syntax of a Matlab
OpenGL command is mostly identical to its C counterpart with a few
small exceptions that are imposed to us by the design of Matlab:

1. Return values are returned in Matlab-style, as left-hand side
   arguments of the calls, instead of being right-hand side arguments as in
   C:

E.g., the C language call glGetIntegerv(GLenum;
becomes params = glGetIntegerv(pname); in Matlab, because ‘params’ is a
return argument of glGetIntegerv.

1. Commands that don’t take arguments don’t have empty braces, because
   Matlab doesn’t allow this:

E.g., the C language call glEnd(); becomes glEnd; in Matlab.

1. All GL, GLU and AGL constants start with prefix GL. instead of GL_
   E.g., GL_RGB becomes GL.RGB, GL_DEPTH_TEST becomes GL.DEPTH_TEST, …

Each subroutine that intends to use GL constants needs to define the
variable GL as global: Example
function myOpenGLSubroutine()
global GL; % Define GL variable as global.
…rest of function implementation…
return;

If you want to use GLU constants, then ‘global GLU’ is also needed.

1. In your main Matlab script or M-File you need to call the function
   InitializeMatlabOpenGL; *before* calling Screen(‘OpenWindow’, …) the
   first time. This command initializes the OpenGL for Matlab toolbox and
   sets up Psychtoolbox to play nicely with Matlab-OpenGL and other OpenGL
   toolboxes. Psychtoolbox will then attach a 24-bit depth buffer (z-buffer)
   and a 8-bit stencil buffer to each onscreen window, so occlusion handling
   works properly when rendering 3D-Stimuli.

Please note that InitializeMatlabOpenGL() allows to optionally set the
‘debuglevel’, the amount of error checking automatically performed during
execution of your scripts. By default, the debug level is set so that
MOGL checks for OpenGL errors after execution of each single OpenGL call!
This is nice for debugging code, but can significantly impact performance
for complex rendering code! Make sure to explicitely set the debuglevel
to ‘0’ once your experiment code is performing as expected, so you can
get higher rendering performance.

Each time after calling a Psychtoolbox Screen() command for 2D drawing,
you need to call Screen(‘BeginOpenGL’, window); to tell PTB that you want
to use OpenGL code to draw into onscreen- or offscreen-window ‘window’,
so PTB can set up the window properly for your OpenGL code. Each time
after you’ve finished drawing with OpenGL commands and you want to draw
with PTB again, you’ll need to call Screen(‘EndOpenGL’, window), so PTB
can switch back to its own drawing engine.

Psychtoolbox provides two Screen subfunctions that allow you to either
use Psychtoolbox textures in your own OpenGL code or to inject your own
self-made OpenGL textures into Psychtoolbox for use with
Screen Drawingcommands. See the Screen online help with…

Screen SetOpenGLTexture?
Screen GetOpenGLTexture?

… for how to use these functions.
This allows you to conveniently upload images into PTB with the usual
img=imread(filename); tex=Screen(‘MakeTexture’, win, img) methods and
then use the texture in your OpenGL drawing code. It also allows you to
access the images of Quicktime movies and images captured by the video
capture functions in your OpenGL code.

We also provide a couple of higher-level functions, implemented as M-Files
to solve common tasks:

* LoadOBJFile() – This implements a simple loader for Alias/Wavefront
OBJ geometry files. Most common 3D graphics applications (e.g, Blender,
Maya, 3D-Studio-Max, …) allow you to export 3D objects and scenes as files
in ASCII-OBJ format. LoadOBJFile() allows you to load such files into
Matlab. See ‘help LoadOBJFile’ for usage info. The loader is pretty
limited at the moment, so read the file carefully to understand its
limitations!

* moglDrawDots3D() – High speed drawing of 3D dots or points, similar to
Screen(‘DrawDots’) for the 2D case.

* moglmorpher() – A high-speed shape rendering and morphing function.
moglmorpher allows to quickly draw (=render) single 3D objects loaded by
LoadOBJFile(). It also allows you to load a collection of shapes and
quickly morph them into each other by linear combination of their shape-
and surface-normal vectors. This is mostly useful for 3D facial animation
and face morphing (face perception studies) and for high-level 3D object
recognition tasks. Have a look at MorphDemo for a nice example of how to
use LoadOBJFile and moglmorpher. This demo was contributed by Quoc, C.
Vuong, MPI for Biological Cybernetics Tuebingen, Germany.

* LoadShaderFromFile() and LoadGLSLProgramFromFiles() – These functions
allow to load OpenGL GL Shading language (GLSL) shader definition files
from the filesystem and to create GLSL shaders from them. When used
properly, one can implement very complex lighting models and a host of
image processing operations directly on the graphics hardware. This can
provide speed gains anywhere from 10 to 1000 times faster than when
executing such algorithms on the CPU. Have a look at GLSLDemo for some
example of how to use GLSL shaders. Use of GLSL shaders requires state of
the art graphics hardware, so if you don’t have a recent graphics adapter
installed in your machine, these demos and functions may fail.

* moglFDF() – A high-speed renderer for “formless dot fields”, random
dot field motion stimuli for the creation of “structure from motion”
stimuli from 3D objects.

More high-level functions will follow…

For demos on how to code OpenGL in Matlab, have a look at the demos in
‘Psychtoolbox/PsychDemos/OpenGL4MatlabDemos/’

Support for 3rd party OpenGL MEX-Files:

You can also code up OpenGL algorithms in the C programming language and
compile them into Matlab-MEX files if you have “need for speed”. Your Mex
files will just contain the mixture of ANSI C code and OpenGL calls, but
no code to setup the window, OpenGL rendering context, or to flip the
front- and backbuffers. Psychtoolbox takes care of setting up OpenGL and
windows for you. You just need to call the InitializeMatlabOpenGL;
function at the beginning of your script and wrap each invocation of your
mex-file into Screen(‘BeginOpenGL’, win) and Screen(‘EndOpenGL’,win)
calls. Use the Screen(‘Flip’, win) command as usual to take care of
stimlulus onset.

If you want to write OpenGL mex-files that are portable across different
operating systems (OS-X, Windows, Linux) then have a look at:
‘Psychtoolbox/PsychOpenGL/MOGL/source’ for how to do this. This folder
contains the source code and Makefiles for our own moglcore mex-file…

KNOWN LIMITATIONS:

If you use many immediate mode OpenGL rendering calls, rendering speed in
Matlab may be significantly lower than when executing the same code from
C or C++. This is the price you’ll have to pay for using Matlab. However,
immediate mode rendering is discouraged even in C for anything but the
most trivial tasks, it’s just that you pay a slightly higher “time
penalty” for doing the wrong thing in Matlab than in C. Well written code
will not cause any significant performance difference to C.

Some OpenGL functions are not yet implemented in the toolbox, because
these functions can’t get automatically generated, so their wrappers need
to be coded manually. Our goal is to provide full support for the
OpenGL-API but finalizing all functions may take some time. Mostly some
of the query-functions - functions that don’t set OpenGL state or execute
some operation, but query the current settings of OpenGL, are missing.

Also, some of the more exotic OpenGL extensions are not yet supported,
especially there is no support for old-style Vertexprograms and
Fragmentprograms, but GLSL vertexshaders and fragmentshaders are a
complete - and easier to use - replacement for these.

Apart from these limitations that will get removed in the future, there
are limitations imposed by your operating system and graphics hardware.
Support for OpenGL functions varies between different graphics hardware,
so if you want to use the latest and greatest OpenGL functions, you’ll
need to buy and install the latest and greatest graphics hardware.

CONTENTS:

* All supported OpenGL low-level functions can be found in the folder
‘Psychtoolbox/PsychOpenGL/MOGL/wrap/’. Functions prefixed with _ are not
yet implemented.

* A number of interesting GLSL shaders for realtime image processing can
be found in ‘Psychtoolbox/PsychOpenGL/PsychGLSLShaders/’. These may only work on
state of the art graphics hardware.

* High-level helper functions (e.g., OBJ file loading, morphing, …) can
be found in ‘Psychtoolbox/PsychOpenGL/’ and its subfolders.

* Demos can be found in ‘Psychtoolbox/PsychDemos/OpenGL4MatlabDemos’

Lot’s of documentation, tutorials, code samples and news about OpenGL can
be found at:

http://www.opengl.org

Enjoy!

Supported functions:

PsychCamIntrinsicsToGLProjectionMatrix - Compute GL_PROJECTION_MATRIX from camera intrinsics.