[Simca-developers] Re: dv compression

[<si...@sm...>]

Hi,

> could you please tell us what did you find out about fw-cameras and
> dv-compression? i'm especially interested to know what colorspace the
> result is in? and what libraries allow to decompress dv? is hw
> decompression possible with some firewire cards?

I don't think hw decompression can be done using firewire card (will look into that) because firewire is just the transport protocol often used to, but not restricted to, dv data. dv compression/decompression is done inside camera or software not by the ieee1394 hardware. 

Basically there are two types of cameras:

1. The normal DV camera. These cameras come in several tastes; from consumer to professional and can use different formats. There is a dv decompression library libdv that is able to decompress into YV12 (planer) and YUY2 (packed). The one to use also depends on the type of camera. Older cameras often used 4:2:0 sampling and so is better to convert to YV12. The newer ones are 4:1:1 and can be upsampled to YUY2 (4:2:2)

2. The DC camera (like the one stephan used for pfom) which is a digital camera. A digital camera differs from a DV camera in that it sends uncompressed image data (which means we don't need decompression of any kind, maybe up or downsampling depending on hardware). The format of DC cameras can differ per type. They are also in relativly cheap webcam versions that still outperform normal webcams (speedwise don't know about image quality). Like the C104T from aplux (http://www.aplux.com). It supports YUV 4:1:1, YUV 4:2:2 and RGB888 30fps (nothing about resolution though). It is possible to use coriander as capture program and loop it to v4l using vloopback. It is then possible to just use v4l module which is easy to write. But I don't think it is to hard to create simca module either.

I'm now looking into porting parts of gstreamer and kino dv code to simca module. Can't test it though because of lack of hardware.

> i'll try to find JPEG or MPEG decompressor that allows that, it'll
> affect the choice of the file formats. You try to figure out if dv
> decompression library allows that. otherwise we'll have to use software
> colorspace conversion.
> 
> Comment: what i previously understood as hw-accelerated colorspace
> conversion is in fact ability to display YUV image, meaning there is no
> RGB result available in the frame buffer :(

I think there should be but it is of no use. There should be plain rgb image during blitting in the physical memory. You should be able to read it out but at that point it is already shown and you probably don't want that.

Cheers,

-- 
Simon de Bakker

\/01|)7	workgroup: http://www.void7.org
personal homepage: http://www.josos.org