Another Great G7/CHDK Benefit!

Started Oct 28, 2007 | Discussions
Abana Senior Member • Posts: 1,603
Another Great G7/CHDK Benefit!

I was playing around with some RAW images I shot today, running them through different converters to see the results. I don't know weather it's the fact that the G7 RAWs aren't specifically supported by the converters or that my limited skills in RAW conversion and editing are the problem, but the results were quite disappointing, and the identical jpg files looked much better. I did, however, discover that turning off NR when shooting jpg makes a huge difference.

In the CHDK RAW Parameters menu, there is an option to turn off noise reduction. Turning off NR applies to RAW and jpg files, even when not shooting RAW. Below are two 160% crops from two identical images shot one after the other. Both were takein using the same settings (see exif) at 400ISO.

The image w/o NR is significantly better, while the image with NR is smeary with a lot of detail lost. These are straight from the G7 with no proccessing other than the crop. Take note of the bulletin board "weave" in the lower right hand corner of both images:

With in-cam noise reduction:

To my eye, the below image (no NR) looks like it has less noise than the image with NR...go figure!

No noise reduction:

It's great to know that even if I can't get decent results from RAW, I can still increase the amount of image detail!

abana

OP Abana Senior Member • Posts: 1,603
Re: Another Great G7/CHDK Benefit!

Today, I took some test shots in daylight (shade) with ISO 400 using the CHDK, shooting in jpg (as before).

Last night's tests (posted above) were taken w/o flash and at ISO 400 with the NR set to on & off, respectively. My test shots today (will be posted upon request) were taken at AUTO, ON, and OFF.

While AUTO seemed to do a better job than ON, still, OFF won hands down for retaining detail and IQ. In fact, in some spots, the in cam NR seemed noisier (as was the case in the posted images). And, when running the image that had no NR through neat image, it still was significantly better than AUTO or ON.

I think this settles it for me. I'll be using the CHDK with NR OFF and performing my own NR in N.I.

abana

Archiver Veteran Member • Posts: 3,708
Re: Another Great G7/CHDK Benefit!

Abana, I agree. Your post about this feature was enough to make me test the CHDK hack, and it works. My G7's images with NR turned off look better to me than those with it turned on, and I may even use this as my default mode for photos.

There is a slight inconvenience as I have to use Play to start the camera, rather than Power, which means there is a much longer lag time between turning it on and taking a shot. I will have to decide what is more important to me - fast response time or better quality images.
--
Archiver - Recording the sights and sounds of life
http://www.flickr.com/photos/archiver/

OP Abana Senior Member • Posts: 1,603
Re: Another Great G7/CHDK Benefit!

Archiver wrote:

Abana, I agree. Your post about this feature was enough to make me
test the CHDK hack, and it works. My G7's images with NR turned off
look better to me than those with it turned on, and I may even use
this as my default mode for photos.

I think the CHDK is under-rated by many who see it as just a way to shoot RAW, and was surprised that this thread didn't generate more responses, given the versatility and potential for greater IQ. I have five 2 GB cards and have the CHDK on everyone (if I have to use them in another cam or just wnat to turn it off, I merely unlock the card and the CHDK doesn't load). It is indeed a bit slower loading, but worth the wait, I think.

There is a slight inconvenience as I have to use Play to start the
camera, rather than Power, which means there is a much longer lag
time between turning it on and taking a shot. I will have to decide
what is more important to me - fast response time or better quality
images.

I've been using the CHDK all the time since it was first ported to the G7. But the NR control is not the only reason to us it. Here is a link to another post describing the DOF calculator, which has greatly improved IQ here:

http://forums.dpreview.com/forums/read.asp?forum=1010&message=24998903

Also the RGB histogram, Zebra mode, scripting...the G7 CHDK truly and significantly advances the functionality of the camera. Add the NR control to these functions and it becomes indispensible, IMO.

Nice gallery, BTW.

abana

OP Abana Senior Member • Posts: 1,603
Re: Another Great G7/CHDK Benefit!

Archiver wrote:

There is a slight inconvenience as I have to use Play to start the
camera, rather than Power,

Sorry...I somehow missed this in my response. I am able to use the power button to power on my G7. At first, it didn't seem to work, but you must only press it once and briefly. Holding it down causes it not to power on. This is unlike pressing 'Play', which must be held down for 3 secs.

My Auto Load CHDK power-up routine is this:

1) press and immediately release the power button (starts in Play Mode). Lingering too long on the Power Button causes the camera not to power up.

2) press the shutter button halfway (this puts it into REC Mode).

This saves 2 or 3 secs over pressing the Play Button.

abana

jose42 Junior Member • Posts: 32
Re: Another Great G7/CHDK Benefit!

Hello abana.

I use the add-on (CHDK) to shot raw in every picture.
The same as you, the NR is off.
At the same time I have a record in jpg. (large & fine).
All pictures are save in this way, using the card reader.
The raw files are converted only in special pictures.
Otherwise I use the jpegs.
The raws are kept like original negatives, untouched.
To convert and edit, I use rawtherapee.

convenient options.
Voltage reading of the battery. For me more precise than percentage.
DOF calculator. With Hyperfocal distance.
Equivalent FL, actual focus distance and actual AV.
I have them displayed all time and found very usefull.

And yes if you start up with the on/off button and then half press the shuter release the difference is minimal.

Regards.Jose

pitquim Contributing Member • Posts: 528
same results with A630 (nt)

.

Ondina Regular Member • Posts: 100
Re: same results with A710 IS

Makes a big difference.

http://www.flickr.com/photos/ondina/

Hansixx Junior Member • Posts: 25
Re: Another Great G7/CHDK Benefit!

Hi there Abana, quite an interesting find, but aside the difference in the upper left dark brown area, where, just as you mentioned the noise seems less with NR off, its obvious that the file with NR on is slightly blurred (by camera shaking).

Thus, i would be very thankful if you could post some other comparison shoot crops.

OP Abana Senior Member • Posts: 1,603
Re: Another Great G7/CHDK Benefit!

Hansixx wrote:

Hi there Abana, quite an interesting find, but aside the difference
in the upper left dark brown area, where, just as you mentioned the
noise seems less with NR off, its obvious that the file with NR on is
slightly blurred (by camera shaking).

Thus, i would be very thankful if you could post some other
comparison shoot crops.

Hi Hansixx,

Here are 3 more crops at 800 ISO. The crops may be slightly different, but all 3 images were shot from a tripod one right after the other. The IQ difference is even more pronounced at the higher ISO. The surprising thing is that (at least@800 ISO) Auto turned in the worst showing here. NR off is still significantly better. I intend to do further testing at different ISOs, focal lengths, etc., but may not be able to do so until the weekend.

The left image is NR off, the right=NR Auto:

This one is NR on:

abana

OP Abana Senior Member • Posts: 1,603
Re: Another Great G7/CHDK Benefit!

jose42 wrote:

Hello abana.

I use the add-on (CHDK) to shot raw in every picture.
The same as you, the NR is off.
At the same time I have a record in jpg. (large & fine).
All pictures are save in this way, using the card reader.
The raw files are converted only in special pictures.
Otherwise I use the jpegs.
The raws are kept like original negatives, untouched.
To convert and edit, I use rawtherapee.

Hi Jose,

I was trying RAW but was disappointed in the results. This was before I started playing with the NR settings, so I will have another go at RAW later in the week with NR off.

Thanks,
abana

Divalent Forum Member • Posts: 99
Very interesting

I posted a note in the CHDK thread to try to get the experts to comment on this.

What is interesting is that you found an effect of a setting that, as far as I can tell from the description at the CHDK Wiki, should not be having any effect. The CHDK documentation describes this feature as preventing the dark frame subtraction, but supposively that only occurs with long exposures ( 0.6 sec and longer).

If that's true, why do you see a difference in images that (from how you describe them) are taken at normal "everyday" exposure times? (Inquiring minds want to know!)

Divalent Forum Member • Posts: 99
Abana and others that see the effect

Could you please post the shutter speed on the image pairs where it is clear that turning the "noise reduction" OFF made a clear difference, along with the camera model? In particular, I'm looking for image pairs where the effect is clear and the shutter speed was on the order of 1/30 or faster.

I ask because (as you can read in my post below) this was thought not to be something that would have any effect on images with shutter speeds about (more or less) 1/2 sec or faster. (But perhaps turning it off also has another unanticipated effect.)

Anyway, just trying to pin this down.

Thanks

Divalent wrote:

I posted a note in the CHDK thread to try to get the experts to
comment on this.

What is interesting is that you found an effect of a setting that, as
far as I can tell from the description at the CHDK Wiki, should not
be having any effect. The CHDK documentation describes this feature
as preventing the dark frame subtraction, but supposively that only
occurs with long exposures ( 0.6 sec and longer).

If that's true, why do you see a difference in images that (from how
you describe them) are taken at normal "everyday" exposure times?
(Inquiring minds want to know!)

Divalent Forum Member • Posts: 99
I tested myself in an S3. It makes sense now.

I tested 800 ISO at 1 sec and at 1/60 sec exposures, 4 shots each of the same scene, two each with "Noiser Reduction" ON and OFF. It clearly is noisier when the NR is ON at both exposures. (And other than worse noise, I saw no difference in the overall intensity or other difference.)

However, I also tested again at 1/60 sec exposure but additionally used the AUTO setting. In that test, I could see no significant difference between the AUTO and OFF settings, but (again) both were better than ON.

So in my tests, it was clear that setting the "NR" to OFF resulted in a lower noise level on the images compared to the ON setting. But it did not improve things compared to the AUTO setting if the image was taken at an exposure where the camera would not normally do the "NR" proceedure.

BTW, the label "noise reduction" is really a misnomer: what this parameter is referring to is the "dark frame" subtraction routine. The purpose of DF subtraction is to remove the time dependent signal on the sensor (a signal that generally is only significant with long exposures, although some individual pixels can be significantly worse than average). You would predict that the DF subtraction process should add noise (and in fact, it does, as everyones tests show), but at least in my S3, at a 1 second exposure it seems to do nothing but add noise.

Thus, setting this to OFF is probably best for most applications (and setting it to ON seems foolish), but for 98% of the shots we take (faster than 1/3 sec exposure), the OFF setting really provides no benefit over AUTO (the native camera mode without CHDK), since the DF subtraction won't be done anyway for those images.

OP Abana Senior Member • Posts: 1,603
Re: I tested myself in an S3. It makes sense now.

Hi Divalent,

Since my last post to this thread, I also have done further testing. My test shots, however, used flash since I have been unable to stop during the daylight hours for the past few days.

At any rate, I took several series of OFF/ON/AUTO shots at different ISOs all from a tripod, all from the same composition, all flash, all 5.6 Av, 1/60 Tv, EV=0, IS off.

800/400/200/100 ISO: all OFF/ON/AUTO sets were compared only withinin same ISO group.

800/400 were easily dinstinguishable. OFF in each ISO was the clear winner. Surprisingly, AUTO was the loser in these groups.

At 200 ISO, the difference becomes less pronounced, but AUTO has a very, very slight edge. Slight enough to have gone either way. ON is the loser.

At ISO 100, AUTO & OFF are indistinguishable. ON is the loser.

Thus, setting this to OFF is probably best for most applications (and
setting it to ON seems foolish), but for 98% of the shots we take
(faster than 1/3 sec exposure), the OFF setting really provides no
benefit over AUTO (the native camera mode without CHDK), since the DF
subtraction won't be done anyway for those images.

Our results were somewhat different. In my tests, Auto was the absolute worst except in the higher ISOs and no better than off in the lower ISOs. Of course, I still have to take daytime shots, but the ISO 100 made it likely that in well lit shots, OFF and AUTO are virtually the same.

For me, that translates to " there is no benefit to using AUTO over Off". Unless my daytime test shots tell me differently, I see no reason not to set my G7 NR to OFF and leave it there.

abana

Divalent Forum Member • Posts: 99
Hmmm.

So at 1/60, 400/800 ISO you found auto to be worse than OFF (and on)?

We do agree that ON is bad.

But the "NR" is not supposed to be active at 1/60 of a second, so in that case AUTO should be equivalent to OFF. Your result suggests that (unlike what I found in my S3 at 800) it did do the DF subtraction in AUTO mode.

However, one variable is flash. I wonder if that has anything to do it, and if so, WHY? Another variable is the camera model.

(My 1/60 shots at 800 were with the aperture wide open and focusing on a lit table lamp, so I was able to get a decent shot without flash)

I'll try flash tomorrow. Could you try flashless at 1/60 (at 800)? (Again, in my hands, AUTO=OFF > ON under these conditions). This is curious.

Cheers,

Divalent

OP Abana Senior Member • Posts: 1,603
ISO 400 & 800 Series~~1 MB Image~~

Divalent wrote:

So at 1/60, 400/800 ISO you found auto to be worse than OFF (and on)?

Upon re-examination, 400 ISO AUTO & ON are very close (sorry for the mis information...too many photos and not enough time), but Auto seems to have the slight edge, Off, however, is the best in all. Here is capture of both 800 & 400 ISO sets. It's rather wide:

OFF/ON/AUTO (as seen in tilte bars).
Top Row: 400 ISO
Bottom Row: 800 ISO

It's worth noting that the above image was captured using Vista's Snip Tool, so additional artifacts may be present (although it looks much the same to my eye).

We do agree that ON is bad.

Yes, ON is bad, but in some cases (higher ISOs), AUTO is significantly worse, as you can (hopefully) see in the 800 ISO series.

But the "NR" is not supposed to be active at 1/60 of a second, so in
that case AUTO should be equivalent to OFF. Your result suggests
that (unlike what I found in my S3 at 800) it did do the DF
subtraction in AUTO mode.

Auto seems prone to introducing artifacts, noise, whatever into the image, at least under the right circumstances.

However, one variable is flash. I wonder if that has anything to do
it, and if so, WHY? Another variable is the camera model.

(My 1/60 shots at 800 were with the aperture wide open and focusing
on a lit table lamp, so I was able to get a decent shot without flash)

I'll try flash tomorrow. Could you try flashless at 1/60 (at 800)?
(Again, in my hands, AUTO=OFF > ON under these conditions). This is
curious.

I may not have the time to take the images tommorow, but I'll try. By thursday if not.

abana

Keyboard shortcuts:
FForum MMy threads