Barrel distortion correction

waldo

238

« on: 31 / December / 2010, 16:08:36 »

Advertisements

Does anyone know where in the firmware the barrel distortion correction is done? I'm trying to see if I can disable that part of the JPEG processing.

I have a need for a JPEG image that retains the pixels in the corner that are removed when Canon corrects for lens distortion. Currently, I can use Raw but I'd rather avoid all the Raw post-processing of white balance, noise reduction, sharpening, etc. just to get those extra corner pixels. I know there is a "fisheye" mode in some cameras, but that seems to take the image after lens distortion correction and just adds additional distortion.

Logged

reyalp

14082

Re: Barrel distortion correction

« Reply #1 on: 31 / December / 2010, 16:34:54 »

Quote from: waldo on 31 / December / 2010, 16:08:36

Does anyone know where in the firmware the barrel distortion correction is done? I'm trying to see if I can disable that part of the JPEG processing.

I have a need for a JPEG image that retains the pixels in the corner that are removed when Canon corrects for lens distortion. Currently, I can use Raw but I'd rather avoid all the Raw post-processing of white balance, noise reduction, sharpening, etc. just to get those extra corner pixels. I know there is a "fisheye" mode in some cameras, but that seems to take the image after lens distortion correction and just adds additional distortion.

This is new territory, I don't recall anyone reporting anything on that (other than the fact that new cameras do correct for barrel distortion)

I would guess the actual correction isn't done with the ARM CPU, but presumably something in the firmware code tells the other side of digic to do it.

Logged

Don't forget what the H stands for.

philmoz

3450

Re: Barrel distortion correction

« Reply #2 on: 31 / December / 2010, 16:45:09 »

Quote from: reyalp on 31 / December / 2010, 16:34:54

I would guess the actual correction isn't done with the ARM CPU, but presumably something in the firmware code tells the other side of digic to do it.

(I was just about to say the same thing.)

On the SX30 the camera is doing a lot of distortion correction at the wide end; but almost none at the tele end.
Perhaps there's some setting being passed to the Digic processor to control the amount of correction - finding it would be interesting.
Has anyone mapped out how the ARM and Digic interact and how processing settings get commincated to the Digic processor?

Regards,
Phil.

Logged

CHDK ports:
sx30is (1.00c, 1.00h, 1.00l, 1.00n & 1.00p)
g12 (1.00c, 1.00e, 1.00f & 1.00g)
sx130is (1.01d & 1.01f)
ixus310hs (1.00a & 1.01a)
sx40hs (1.00d, 1.00g & 1.00i)
g1x (1.00e, 1.00f & 1.00g)
g5x (1.00c, 1.01a, 1.01b)
g7x2 (1.01a, 1.01b, 1.10b)

waldo

238

Re: Barrel distortion correction

« Reply #3 on: 31 / December / 2010, 16:49:24 »

Quote from: reyalp on 31 / December / 2010, 16:34:54

I would guess the actual correction isn't done with the ARM CPU, but presumably something in the firmware code tells the other side of digic to do it.

I would also guess that they did it that way. Having explored capt_seq task recently, I know that the JPEG processing is done heavily in hardware, but is started within the capt_seq task after the image capture. That is why cap_seq_hook_raw_here contains a wait loop - it halts the capt_seq task and prevents starting the JPEG processing in hardware until things like raw develop have had a chance to finish. I'm hoping that something similar could be done for lens distortion.

Logged

reyalp

14082

Re: Barrel distortion correction

« Reply #4 on: 31 / December / 2010, 17:26:43 »

Quote from: philmoz on 31 / December / 2010, 16:45:09

Has anyone mapped out how the ARM and Digic interact and how processing settings get commincated to the Digic processor?

My guess is EngDrv* is relevant.

There are also some chunks of stuff that appears to be non-ARM code related to jpeg processing. In older cameras (e.g. a540), the strings like
"jp52_firmware build#0014 2005/01/31 Akihiro Uehara, Satoshi Naito, Yukio Chiba"
"launcher.s rev-1.0 2001/07/17 author Satoshi Naitod"
are associated with these chunks.

Similar chunks exist in newer cameras, minus the strings.

Quote from: waldo

That is why cap_seq_hook_raw_here contains a wait loop - it halts the capt_seq task and prevents starting the JPEG processing in hardware until things like raw develop have had a chance to finish.

This just blocks capt_seq while our raw code runs in spytask.

FWIW, I believe there is at least one another task related to jpeg developing, maybe DevelopModule.

« Last Edit: 31 / December / 2010, 17:33:21 by reyalp »

Logged

Don't forget what the H stands for.

waldo

238

Re: Barrel distortion correction

« Reply #5 on: 31 / December / 2010, 19:00:59 »

Here's what I have explored so far:

Capt_seq_task:
I've looked at all 35 jumptable entries and none seem to do or to start the actual jpeg processing

SsDvlpSeq.c:
Inside DvlpSeqTask, one of the sub calls is to a function that gets propcases for jpeg resolution, quality, date stamp, etc that may be promising. So far, I haven't found a propcase that turns off distortion correction (wishful thinking!). I've tried patching this task but I haven't found anything interesting so far.
There may be a table somewhere with lens geometry corrections, but patching a table would probably require patching the function that uses the table.

« Last Edit: 31 / December / 2010, 19:05:14 by waldo »

Logged

Bernd R

430

Re: Barrel distortion correction

« Reply #6 on: 19 / January / 2011, 10:09:28 »

have you maybe find during your tests what code do the Canon Noise reduction and sharpening ?

I mean not dark frame subtraction.
because cams have lots texture loss, maybe its possible to do less denoise or at least switch off complete and do all sharpen and denoise on PC.

sure there is raw, but RAW is slow and on IX1000 it not work

« Last Edit: 19 / January / 2011, 10:11:13 by Bernd R »

Logged

Ixus 1000 HS