H3A module on omap4 iss

H3A module on omap4 iss

[Sriram V]

Hi,

I am planning to work on getting the H3A support (Auto White balance,
Auto Exposure and Auto Focus Support) for OMAP4 ISS.

I wanted to check if anyone is working to get this module working.

Is OMAP3 H3A similar or same as OMAP4 H3A. Can i make use of the OMAP3
code base for h3a or how complex is this activity?

Also, i wanted to how how to test the v4l resizer driver in omap4iss?

Any help is highly appreciated.

-- 
Regards,
Sriram

Re: H3A module on omap4 iss

[Laurent Pinchart]

Hi Sriram,

On Wednesday 20 August 2014 11:55:12 Sriram V wrote:
> Hi,
> 
> I am planning to work on getting the H3A support (Auto White balance,
> Auto Exposure and Auto Focus Support) for OMAP4 ISS.

That would be great !

> I wanted to check if anyone is working to get this module working.

Not to my knowledge. I'm not (at least for now, but I could need H3A support 
at some point in the future).

> Is OMAP3 H3A similar or same as OMAP4 H3A. Can i make use of the OMAP3 code
> base for h3a or how complex is this activity?

The AF and AE engines look pretty similar indeed. It would be nice if the 
OMAP3 ISP AF and AE code could be shared with the OMAP4 ISS driver. That would 
require some refactoring though. One thing we're missing is standardization of 
H3A APIs in V4L2, I'd like to see that happen.

> Also, i wanted to how how to test the v4l resizer driver in omap4iss?

Here's the full pipeline configuration I use. Please note that the OMAP4 ISS 
resizer implementation doesn't support resizing yet, it just hardcodes the 
resizer ratio to 1:1.

IFMT="SBGGR10"
ISIZE="1920x1084"
OFMT="UYVY"
OSIZE="1920x1080"
CFMT="YUYV1_5X8"

media-ctl -l '"OMAP4 ISS CSI2a":1 -> "OMAP4 ISS ISP IPIPEIF":0 [1]'
media-ctl -l '"OMAP4 ISS ISP IPIPEIF":2 -> "OMAP4 ISS ISP IPIPE":0 [1]'
media-ctl -l '"OMAP4 ISS ISP IPIPE":1 -> "OMAP4 ISS ISP resizer":0 [1]'
media-ctl -l '"OMAP4 ISS ISP resizer":1 -> "OMAP4 ISS ISP resizer a output":0 
[1]'

media-ctl -V "\"sensor\":0 [fmt:$IFMT/$ISIZE]"
media-ctl -V "\"OMAP4 ISS CSI2a\":1 [fmt:$IFMT/$ISIZE]"
media-ctl -V "\"OMAP4 ISS ISP IPIPEIF\":0 [fmt:$IFMT/$ISIZE]"
media-ctl -V "\"OMAP4 ISS ISP IPIPE\":0 [fmt:$IFMT/$ISIZE]"
media-ctl -V "\"OMAP4 ISS ISP IPIPE\":1 [fmt:$OFMT/$OSIZE]"
media-ctl -V "\"OMAP4 ISS ISP resizer\":0 [fmt:$OFMT/$OSIZE]"
media-ctl -V "\"OMAP4 ISS ISP resizer\":1 [fmt:$CFMT/$OSIZE]"

-- 
Regards,

Laurent Pinchart

Re: H3A module on omap4 iss

[Sriram V]

Thanks for your response.

Going through the Datasheet & omap3isp code. what i understand is
The code enables the H3A engine for AF, AE and AWB.

Now, the H3A produces co-efficients which need to be used for achieving
actual AF, AE & AWB.

Now, I could not find any references where these co-efficents are used to
do AF, AE & AWB.

My specific question is now if i need to do AF - Do you know any
formula or algorithm
which can make use of these co-coefficients and control the sensor motor.

Similarly - How AWB is done with the help of these co-efficients when
AWB is enabled?

Correct me if i am wrong.

Thanks.


On Wed, Aug 20, 2014 at 8:02 PM, Laurent Pinchart
<laurent.pinchart@ideasonboard.com> wrote:
> Hi Sriram,
>
> On Wednesday 20 August 2014 11:55:12 Sriram V wrote:
>> Hi,
>>
>> I am planning to work on getting the H3A support (Auto White balance,
>> Auto Exposure and Auto Focus Support) for OMAP4 ISS.
>
> That would be great !
>
>> I wanted to check if anyone is working to get this module working.
>
> Not to my knowledge. I'm not (at least for now, but I could need H3A support
> at some point in the future).
>
>> Is OMAP3 H3A similar or same as OMAP4 H3A. Can i make use of the OMAP3 code
>> base for h3a or how complex is this activity?
>
> The AF and AE engines look pretty similar indeed. It would be nice if the
> OMAP3 ISP AF and AE code could be shared with the OMAP4 ISS driver. That would
> require some refactoring though. One thing we're missing is standardization of
> H3A APIs in V4L2, I'd like to see that happen.
>
>> Also, i wanted to how how to test the v4l resizer driver in omap4iss?
>
> Here's the full pipeline configuration I use. Please note that the OMAP4 ISS
> resizer implementation doesn't support resizing yet, it just hardcodes the
> resizer ratio to 1:1.
>
> IFMT="SBGGR10"
> ISIZE="1920x1084"
> OFMT="UYVY"
> OSIZE="1920x1080"
> CFMT="YUYV1_5X8"
>
> media-ctl -l '"OMAP4 ISS CSI2a":1 -> "OMAP4 ISS ISP IPIPEIF":0 [1]'
> media-ctl -l '"OMAP4 ISS ISP IPIPEIF":2 -> "OMAP4 ISS ISP IPIPE":0 [1]'
> media-ctl -l '"OMAP4 ISS ISP IPIPE":1 -> "OMAP4 ISS ISP resizer":0 [1]'
> media-ctl -l '"OMAP4 ISS ISP resizer":1 -> "OMAP4 ISS ISP resizer a output":0
> [1]'
>
> media-ctl -V "\"sensor\":0 [fmt:$IFMT/$ISIZE]"
> media-ctl -V "\"OMAP4 ISS CSI2a\":1 [fmt:$IFMT/$ISIZE]"
> media-ctl -V "\"OMAP4 ISS ISP IPIPEIF\":0 [fmt:$IFMT/$ISIZE]"
> media-ctl -V "\"OMAP4 ISS ISP IPIPE\":0 [fmt:$IFMT/$ISIZE]"
> media-ctl -V "\"OMAP4 ISS ISP IPIPE\":1 [fmt:$OFMT/$OSIZE]"
> media-ctl -V "\"OMAP4 ISS ISP resizer\":0 [fmt:$OFMT/$OSIZE]"
> media-ctl -V "\"OMAP4 ISS ISP resizer\":1 [fmt:$CFMT/$OSIZE]"
>
> --
> Regards,
>
> Laurent Pinchart
>



-- 
Regards,
Sriram

Re: H3A module on omap4 iss

[Laurent Pinchart]

Hi Sriram,

On Wednesday 20 August 2014 21:01:30 Sriram V wrote:
> Thanks for your response.
> 
> Going through the Datasheet & omap3isp code. what i understand is
> The code enables the H3A engine for AF, AE and AWB.
> 
> Now, the H3A produces co-efficients which need to be used for achieving
> actual AF, AE & AWB.
> 
> Now, I could not find any references where these co-efficents are used to
> do AF, AE & AWB.
> 
> My specific question is now if i need to do AF - Do you know any formula or
> algorithm which can make use of these co-coefficients and control the sensor
> motor.

I haven't worked with AF personally, I can't really help you.

> Similarly - How AWB is done with the help of these co-efficients when
> AWB is enabled?

I've implemented simple AE and AWB code for the OMAP3 ISP. You can find it at 
http://git.ideasonboard.org/omap3-isp-live.git (look at the iq_aewb_process 
function in iq.c).

The code should be considered as a proof of concept, in a real system you will 
very likely want to implement more complex AE and AWB algorithms. I'd be happy 
to receive patches that make my trivial implementation smarter :-)

-- 
Regards,

Laurent Pinchart