16 Bit per component RBG

[khan]

Is it possible to use/render 16 Bit per pixel (48 bits per component) RGB images with OpenGl? I have a source RGB image which is 48 bits per pixel and I would like to render some graphics into the image and then store it out as a 48 bit per pixel image. What other pixel depths can be used for outputting graphics? What are the lkimitations? Any references on this issue will be greatly appreciated.

[Robbo]

With most consumer level gfx cards at present you won't be able to get any more than a 32 bit pixel format. 64 bit should be with us soon but I guess that doesn't help you at all! My advice would be to composite your image by generating your overlaid image in 32bpp and pasting that into your scene.

[Azdo]

Originally posted by Robbo:

64 bit should be with us soon.

I'm just curious: What are the advantages of 64 bit colour?

[khan]

Originally posted by Azdo:
I'm just curious: What are the advantages of 64 bit colour?

For certain types of images, medical images for example, 8 bits per pixel dos not provide enough dynamic range to represent all the brightness information present in an image. It is in these situations that one has to use a larger nuymber of bits per pixel component.

[tcobbs]

Originally posted by khan:
For certain types of images, medical images for example, 8 bits per pixel dos not provide enough dynamic range to represent all the brightness information present in an image. It is in these situations that one has to use a larger nuymber of bits per pixel component.

I think the more important driving factor for consumer-level hardware is the fact that multi-pass algorithms lead to accumulated round-off error, which results in banding in the final result. The more passes you have, the more round-off error is likely to accumulate, and the more banding you will have in the final image.