Is this even possible?
Unless, i would be able to remove the aliasing by some other way.
I know OpenGL had some option to enable antialiasing with GL_POLYGON_SMOOTH back in the days,
i seems that nowadays most graphic-cards could only do Antialiasing with the multisample option.
And in my case the graphic card dont support multisample, so what other way could one do to remove
aliasing artifacts and also without the use of shaders?
I've attached an image of some cubes joined together, they seem to give some aliasing, that is the edges of the
polygons aint placed evenly together when they render. Is there a way to fix this?
[edit: im looking into the depth bias]
Antialiasing without multisampling?
Without the use of any shaders this could get quite tricky
What kind of card are you using exactly if it doesn't support multisampling?
There are some very decent alternatives to MSAA which use edge detection algorithms (like MLAA), but I don't see a way to implement those without the use of shaders
What kind of card are you using exactly if it doesn't support multisampling?
There are some very decent alternatives to MSAA which use edge detection algorithms (like MLAA), but I don't see a way to implement those without the use of shaders
Author
Without the use of any shaders this could get quite tricky
What kind of card are you using exactly if it doesn't support multisampling?
There are some very decent alternatives to MSAA which use edge detection algorithms (like MLAA), but I don't see a way to implement those without the use of shaders
Actually its a strange hybrid of Intel® GMA X4500hd, i think.
some gl-extensions i can find here: cantiga
and its gpu integrated in laptop: mobile esprimo v6535
this card doesnt support very much use of shaders. only ehm 1.10 i know of that is 100% supported.
shader-versions over that is more or less not-supported.
(edit: i dont have much experience in coding shaders, so it would be a bit daunting task to do anyway).
You can do supersampling by rendering to a target that is 2x or 4x in resolution and then using standard texture sampling as a means to smooth detail when downsampling.
You can do supersampling by rendering to a target that is 2x or 4x in resolution and then using standard texture sampling as a means to smooth detail when downsampling.
Seeing as he is using a rather weak integrated card this could get rather expensive even for low resolutions
Author
You can do supersampling by rendering to a target that is 2x or 4x in resolution and then using standard texture sampling as a means to smooth detail when downsampling.
hmm, yea, good idea. ill try that out.
by the way what i've discovered, which i didnt mention before was that
it just doesnt seem right that there would be white jaggies on the parts that are dark on that image.
i could try out changing the depth bias for each cube though. maybe that would work.
so my next question is this: would it be possible to retrieve the z-buffer value from each fragment and make the change per fragment in the z-bias
to remove those artifacts?
actually, when i come to think about it, i have a hunch that antialiasing the cubes or the whole back/front-buffer with that image would probably not
reduce the white stripes entirely, but i might be wrong. it's one thing i would try to figure out in a while.
thanks for the replies.
Author
Seeing as he is using a rather weak integrated card this could get rather expensive even for low resolutions
Do you mean that the GetFrontBufferData (or BackBuffer for that matter) is too expensive for that, per frame?
I havent really tested it, but you might be right. it probably would not work. Since i want this in a reasonable framerate.
[quote name='Radikalizm' timestamp='1311090297' post='4837448']
Seeing as he is using a rather weak integrated card this could get rather expensive even for low resolutions
Do you mean that the GetFrontBufferData (or BackBuffer for that matter) is too expensive for that, per frame?
I havent really tested it, but you might be right. it probably would not work. Since i want this in a reasonable framerate.
[/quote]
It mostly depends on how populated your scene will be, try to find out how many ms it takes to render 1 frame right now to get an estimate of how much time it will take to render to higher resolutions
If your scene will be simple without too many independent entities (ie. without too many draw calls, polygons, 'complicated' shading, etc.) you should be fine
I'd just try the suggested method and see how it works out
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
