Re: Lower-cost IIDC cameras

Reply to the mail from Andrew Hogue (hogue@cs.yorku.ca):

> Hi again Arne,
>
> I was just looking at the Datasheet for the TI chipset and the errata for
> the chip. I quote the following paragraph:
>
> <begin quote>
> 5. color_bw Bit Cannot Be Configured for Monochrome Operation
>
> Problem:
>
> The camera works properly when the color_bw bit in the VIDEO_OPTIONS_CNFG
> register (0F22h) is set to 1, indicating that the TSB15LV01 is to assume
> the CCD is a color CCD and process it accordingly. However, if set to 0,
> the resulting image has a cross-hatched appearance.
>
> Workaround:
>
> As a result of this problem, TSB15LV01 cannot be used with a monochrome
> sensor. It only works with the Sony or Sharp color sensors. If a
> monochrome image is desired, the monochrome transfer mode can be used.
> Note that with this setting, the image from the sensor is processed as a
> color image including interpolation of a presupposed Bayer pattern
> and then only the luminance data is transmitted to the host.
> <end quote>
>
> You stated:
> > IIRC you need to program the following register:
> > 0xF20 VIDEO_OPTIONS_CNFG
> > If you clear bit 30: color_bw, you should get the bayer pattern.
>
> Aren't these the same things? The documentation says that if this bit is
> set to "0" then the image has a "cross-hatched" appearance. Is this
> simply the bayer pattern or is it some sort of interpolated version of the
> bayer pattern?

This errata seems to be recently added - I have not noticed it previously ( the change date says: 31.Aug.2004 ).

If you set this bit to 0, you will get the bayer pattern. If you are using a monochrome sensor, you will also need to make sure that the white balance registers are set to "64", otherwise you will get what I guess they meen by "cross-hatched" image.

 -Arne

>
> Thanks,
>
> -- Andrew
>
>
>
> On Sat, 13 Nov 2004, Arne Caspari wrote:
>
> > Andrew Hogue wrote:
> >
> > >Hi Arne,
> > >
> > >I would love to hear any advice on 'tuning' the TI chipset you may have.
> > >
> > >
> > >
> > >>If you are not happy with the debayering code of the TI chipset, you can
> > >>tell the chip to use a monochrome CCD ( get the spec of the TI TSB15Lv01
> > >>). This will give you the raw bayer pattern of the CCD. Then you can do
> > >>the bayer processing on the host PC and use a better algorithm ( you
> > >>need to do another workaround for a bug in the chipset then but I can
> > >>give you details if you ask for it ;-) ).
> > >>
> > >>
> > >
> > >I would be grateful for any information you would be willing to provide
> > >about this. It would be wonderful to have access to the raw bayer image
> > >so I could do the processing my self... some of the work I do requires
> > >offline computation anyways.
> > >
> > >
> >
> > Andrew,
> >
> > At first, you should go to www.ti.com and get the datasheet of the
> > "TSB15LV01" ( do a part search on this part ID ).
> >
> > I assume you already know about debayering algorithms.
> >
> > IIRC you need to program the following register:
> > 0xF20 VIDEO_OPTIONS_CNFG
> > If you clear bit 30: color_bw, you should get the bayer pattern.
> >
> > Now you should set "WhiteBalance", Saturation etc to the value 64 which
> > should be the "neutral" values for this DSP.
> >
> > Then you might come across the next issue: The DSP will change the
> > starting row of the output randomly on every startup. The solution I
> > found for this is to turn on the Test-Pattern using the register:
> > 0xF04 TEST_CNFG
> >
> > Locate eg. the red bar of this test pattern. In bayer mode, check if you
> > get red pixels in the location where the red bar is. If it is not, you
> > need to alter your debayer code to compensate for the changed starting row.
> >
> > I hope this information helps. Otherwise, ask ;-)
> >
> >
> > -Arne
> >
> >
> > -------------------------------------------------------
> > This SF.Net email is sponsored by: InterSystems CACHE
> > FREE OODBMS DOWNLOAD - A multidimensional database that combines
> > robust object and relational technologies, making it a perfect match
> > for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8
> > _______________________________________________
> > mailing list Linux1394-user@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/linux1394-user
> >

-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8
_______________________________________________
mailing list Linux1394-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux1394-user
Received on Mon Nov 15 15:04:17 2004