On Sun, 2002-05-19 at 16:24, Steven Ellis wrote: > I have been taking a poke around the v4l sources and I > noticed one driver doing the appropriate conversion. I > just wanted to get my head around some of the maths. The current idea is that the driver MUST NOT do such conversions, unless they are absolutely necessary (when the camera datastream, for example, is too different from any standard format). The color conversion, if needed, should be done by the v4l application (or library) - it will be of higher quality and better optimized for the CPU. The driver code must be kept to a minimum. Alan's somewhat minimalistic approach is that the driver just dumps whatever it gets from the camera into userspace, and it will be up to the application to sort it out. Dmitri
Attachment:
signature.asc
Description: This is a digitally signed message part