Re: [libdc] reading mono16?

[Leif A. <le...@ad...>]

On 09/03/15 02:58, Damien Douxchamps wrote:
> Hi Leif,
>
> On Mon, 2015-03-09 at 01:44 +0100, Leif Adelw wrote:
>> Dear fellow firewires,
>> I have a set of AVT cameras that I need linux access to. I've been
>> trying to fetch 16 bit images and dumping them into opencv, cimg or pgm
>> files. So far I've only been able to extract 256 grayscales from frames
>> irregardless of "color coding". Is there anyone that can throw an
>> example at me (I've looked at and compiled the grab_*.c and avt_*
>> examples from libdc1394)?
> The "color coding" setting is only available for customizable image
> formats ("format7", see the definition of dc1394video_mode_t in
> dc1394/types.h). It won't have any effect if you are streaming a
> "standard" image format. You could try using these "standard" image
> format first such as DC1394_VIDEO_MODE_1600x1200_MONO16 (or another
> resolution supported by your camera).
>
>>    BUT when dumping raw files with coriander I clearly see a difference
>> if choosing a 16 bit format.
> You see a difference but it still doesn't work, right?
>
>> Reading 8 bit with e.g., {octave,
>> a=fread(fid,[640 480],'uint8');} is fine but 16 bit raw file with
>> {octave,a=fread(fid,[640 480],'uint16');} yield "nonsense". Is this
>> rather a 14bit x 640 x 480 stream that needs to be dealt with, i.e.
>> rather than beeing packed in 16 bit containers?
> There is no such packing for 14 bits formats; it is sent as a 16bit
> integer. However, for historical reasons some manufacturers populate the
> LSBs while others populate the MSBs. You can verify this if you are
> using a format7 format: in the first case they should set the data depth
> to 14, in the second case it should be 16. You can get this data depth
> with a call to dc1394_format7_get_data_depth().
>
>
Thanks, Damien
I just read the 
http://damien.douxchamps.net/ieee1394/libdc1394/iidc/IIDC_2.0.pdf doc 
(p. 100). Seems as if (reading coriander 16 bit raw with octave):
---
  fid=fopen('test-0000000004.raw','rb');
  a=uint16(fread(fid,[640 480],'uint16'));  %one of the "standard" 
format reported by mensioned camera
  b=swapbytes(a);
---
,results in a correct 16 bit array/image.

Moreover, dc1394_format7_get_data_depth() report 14 bits in my case, 
hence LSB. I'll alter my code and report back after some hours of sleep.
Thanks, again

Best Regards,
Leif Adelöw