K5 and high ISO in the real world (our kitchen)

The moire is still quite visible at 100%

The meandering pattern is still left - and is easily visible at 200%

The problem is still there. Probably decreased to invisible far all but the biggest prints or smallest crops.

There are other algorithms that dont do it.

BTW - take a look at the white wall to the right. It has a rather strange repeatable pattern. Hard to know - but it is probably not there. Its probably totally invented.

No matter if it is invisible in most images. It decreases the potential uses of the camera. If you use another converter you will get a better result.

--
Roland

support http://www.openraw.org/
(Sleeping - so the need to support it is even higher)

X3F tools : http://www.proxel.se/x3f.html

HI There
Well - I actually think that that DNG file is 'already damaged'
it's been tried in various converters.
Is it, for example, straight from the camera.

That kind of MAZE pattern is something I've seen before from lightroom and bibble and C1, but not ever from Aperture - was the image perhaps converted using DNG converter?

I'm surprised at it, that's all

all the best
--
Jono Slack
http://www.slack.co.uk

Jono Slack said:

Well - I actually think that that DNG file is 'already damaged'
it's been tried in various converters.
Is it, for example, straight from the camera.

That kind of MAZE pattern is something I've seen before from lightroom and bibble and C1, but not ever from Aperture - was the image perhaps converted using DNG converter?

I'm surprised at it, that's all

Click to expand...

Jono, yes, the DNG as Manu has distributed it has been created by the free Adobe DNG Converter application from a PEF file, as is obvious in that the K-m/K2000D cannot create compressed DNG's as this is and that it contains the MakerNotes information inside the PrivateDNG tag preceded by a Adobe header rather than the straight "AOC" Pentax header as if it had been converted from an in-camera generated uncompressed DNG file.

However, the conversion to DNG by the Adobe application does not change the raw data as to making colour moiré worse nor introducing "mazing" demosiacing artefacts. These two effects are solely there due to the weak Anti Aliasing (AA) filter of the K-m/K2000D (which is common to the weak AA filters of the K10D and K200D that share the same 10 MP CCD sensor, not that the AA filter is directly related to the sensor), and the demosiacing of the raw converter, respectively.

As has been said, demosiacing algorithms that extrapolate based on vectors such as that used by Adobe Camera Raw (ACR) and most other commercial products generally produce more side effect artefacts such as this "mazing" and others at the same time as they can extract the maximum in detail resolution. In comparison, averaging types of demosiacing algorithms tend not to produce these types of artefacts but aren't the ultimate in extracting the maximum detail resolution. In effect, extrapolating vector-based algorithms are trying to "connect the dots" of the patterns seen in the raw data, and in this case are getting confused by the "false detail" provided by the aliased frequency data; other sources of "false detail" are high levels of random noise as in high ISO sensitivity images, which can cause a similar "mazing" effect.

For this image, you are likely seeing more of this "mazing" even in Aperture as you have not before seen so much "false detail" as in this aliased example.

Regards, GordonBGood

Roland Karlsson said:

No matter if it is invisible in most images. It decreases the potential uses of the camera. If you use another converter you will get a better result.

Click to expand...

Roland, you are correct that if one uses a different raw converter based on a different demosiacing algorithm one with get a different result; however it may not be much better as to less "mazing" artefacts unless one uses an algorithm that averages rather than extrapolates data at some cost to maximum detail resolution.

However, the aliasing colour moiré will only be reduced by techniques that either smear the moiré patterns away at a great loss of detail in those areas or which apply the green channel detail pattern to the red and blue colour data, whether done manually in post processing or automatically by the raw conversion algorithm. This is because the aliasing detail is "real" as far as the sensor is concerned, and the processing is required to turn what it "saw" into what we would rather it received.

Regards, GordonBGood

GordonBGood said:

Jono, yes, the DNG as Manu has distributed it has been created by the free Adobe DNG Converter application from a PEF file,

Click to expand...

Hello Gordon, you guessed right. The K-m was not able to produce native compressed DNG like the later cameras (starting from K-7). So I shot compressed PEF then converted to DNG. This was converted from Lightroom (which probably uses the same engine as the free Adobe DNG converter). It comes to my mind that the DNG converter offer the option to convert to linearized (demosaiced) DNG but I think (and hope) it's not the case here as I always prefer to keep the raw untouched for later processing with a possibly better demosaicing algorithm. The downside of converting to DNG is that AFAIK it's not compatible anymore with the Pentax raw utility (as Roland Karlsson found out).

Also thanks for your explanations. As I understand the maze is created by the software that is trying to connect the dots. And the moire a side effect of the Bayer pattern with a too light AA filter.

It would be interesting (for me anyway) to retake the same scene with the same lens but this time with the K-5.

--



http://flickriver.com/photos/ensh/popular-interesting/
http://www.flickr.com/photos/ensh/
PPG: http://bit.ly/cQhegL

HI Gordon
Thanks for that - I was half way there.

I'd simply never seen Mazing in Aperture (and I've been using cameras with no AA filter for several years).

The moire is easy to deal with (at least, it's easy to deal with in Aperture). That lot was fairly comprehensive, and would have taken at least 2 minutes (I only gave it one ).

It seems to me that the only fly in the K5's ointment is, in fact, a rather heavy AA filter, I can live with it, but I might be tempted to do something about it if I really fall in love with the camera.

Manu - I could do another try so that there was no moire, but it's not going to impress anyone!

all the best
--
Jono Slack
http://www.slack.co.uk

Hi Gordon

GordonBGood said:

However, the aliasing colour moiré will only be reduced by techniques that either smear the moiré patterns away at a great loss of detail in those areas or which apply the green channel detail pattern to the red and blue colour data, whether done manually in post processing or automatically by the raw conversion algorithm. This is because the aliasing detail is "real" as far as the sensor is concerned, and the processing is required to turn what it "saw" into what we would rather it received.

Click to expand...

Hi there

actually, it can be dealt with quite easily with some software. The tool in Aperture does a fairly good job, but you can almost completely deal with it using the brush tools, without losing any of your nice MAZE detail

This is because the moire is coloured differently, so you can very easily selectively desaturate the colour in that area (usually yellow and blue) by brushing it in.

To be honest, out of 50,000 shots (and more) with M8s and M9s, I don't think I've needed to do anything with it on more than a dozen occasions - but it can be done, and it really isn't that tough.

So - personally, I rather regret the strength of the AA filter on the K5, but to be realistic, it only really shows up when pixel peeping at distant vegetation, but it's there nevertheless.

Seems to me that all the camera makers seem to vacillate between overdoing the AA filter . . .. and then underdoing it and having complaints about moire!

all the best
--
Jono Slack
http://www.slack.co.uk

ManuH said:

GordonBGood said:

Jono, yes, the DNG as Manu has distributed it has been created by the free Adobe DNG Converter application from a PEF file,

Click to expand...

Hello Gordon, you guessed right. The K-m was not able to produce native compressed DNG like the later cameras (starting from K-7). So I shot compressed PEF then converted to DNG. This was converted from Lightroom (which probably uses the same engine as the free Adobe DNG converter). It comes to my mind that the DNG converter offer the option to convert to linearized (demosaiced) DNG but I think (and hope) it's not the case here as I always prefer to keep the raw untouched for later processing with a possibly better demosaicing algorithm. The downside of converting to DNG is that AFAIK it's not compatible anymore with the Pentax raw utility (as Roland Karlsson found out).

Also thanks for your explanations. As I understand the maze is created by the software that is trying to connect the dots. And the moire a side effect of the Bayer pattern with a too light AA filter.

It would be interesting (for me anyway) to retake the same scene with the same lens but this time with the K-5.

Click to expand...

Hi Manu

I'm afraid I'm a Pentax newbie, so I didn't realise that the K-m didn't produce DNG files (one of the attractions of the K5 is that it does).

As for taking the same shot with the K5 - go for it . . . . I guess you'll be regretting the lack of that nice detail!

all the best
--
Jono Slack
http://www.slack.co.uk

How can you "do something" about a heavy AA filter? I have an Olympus E620 that I love except for that AA filter and would like it removed. Does anyone do that? I wondered about the IR conversion people -- Do they know how to remove an AA filter?
--
http://www.flickr.com/photos/75673106@N00/

Jono Slack said:

So - personally, I rather regret the strength of the AA filter on the K5, but to be realistic, it only really shows up when pixel peeping at distant vegetation, but it's there nevertheless.

Click to expand...

I think for a landscapist a weak AA filter is probably preferable as they will not encounter much situations where the moiré effect will show up. OTOH if you like shooting building or anything with fine straight lines or geometric patterns, a weak AA filter can get very annoying to fix.

I don't think a weak AA filter gives much more real details. It just looks sharper but you can get that look easily with a good sharpening tool.

--



http://flickriver.com/photos/ensh/popular-interesting/
http://www.flickr.com/photos/ensh/
PPG: http://bit.ly/cQhegL

Jono Slack said:

Not sure how big you were planning to print this, but I would have thought that it would be pretty big before you needed to worry about these artifacts, and as a web image I don't think they signify at all!

Click to expand...

Thanks for trying. It's better but you still see strange colors albeit this time you would have to look closely to notice them on a decently sized print.

To get back to topic, I don't think the K-5 AA filter is too strong. It's just right IMHO. I'm very pleased by the amount of details I see on my first pictures with this camera. You would have to print very very big to notice anything about the AA filter difference, especially if you use the right amount of sharpening.

Actually, because of sharpening, I would bet that nobody would be able to see the difference between a camera with a strong AA filter but sharpened and one with no filter. At least on normal viewing (web or print). It may be only visible at 100% on the screen. OTOH moiré is visible even at small size, unless you spend time trying to fix it.

--



http://flickriver.com/photos/ensh/popular-interesting/
http://www.flickr.com/photos/ensh/
PPG: http://bit.ly/cQhegL

GordonBGood said:

Roland, you are correct that if one uses a different raw converter based on a different demosiacing algorithm one with get a different result; however it may not be much better as to less "mazing" artefacts unless one uses an algorithm that averages rather than extrapolates data at some cost to maximum detail resolution.

Click to expand...

I have, as I wrote elsewhere, made some experiments with writing diverse demosaicing code. And I know why the meandering patterns appear. Its because you are trying to be clever and use an adapting algorithm that adapts too fast. For every pixel you try to find its optimal sharpness.

If you instead dont adapt - then you will get an averaging algorithm that blurs the image. Thats actually more true to the image - but it looks dull.

GordonBGood said:

However, the aliasing colour moiré will only be reduced by techniques that either smear the moiré patterns away at a great loss of detail in those areas or which apply the green channel detail pattern to the red and blue colour data, whether done manually in post processing or automatically by the raw conversion algorithm. This is because the aliasing detail is "real" as far as the sensor is concerned, and the processing is required to turn what it "saw" into what we would rather it received.

Click to expand...

Yepp .. the aliasing is due to too little anti aliasing

The meandering patterns although is probably not. Its due to to fast adaptation. Of course - you can remove the meandering patterns by having a very strong AA filter as there are no sharp edges to meander on. But ... the image does not have to be aliased to get them.

--
Roland

support http://www.openraw.org/
(Sleeping - so the need to support it is even higher)

X3F tools : http://www.proxel.se/x3f.html

Jono Slack said:

2. The K5 shots are fine at normal ISO (but they don't really compete with those from the M9). Above 800 ISO they are so much better than the M9 that it isn't funny.

Click to expand...

Hi Jono,

I haven't seen many people who have both Pentax and Leica systems, so I'm very interested in your opinion.

Could you describe what it is that you prefer about the images from your M9 at normal ISO? Is it just the rendering from the Leica lenses or are you seeing something else?

Thanks!

--
Jeff Kott

Jono Slack said:

Hi Manu

I'm afraid I'm a Pentax newbie, so I didn't realise that the K-m didn't produce DNG files (one of the attractions of the K5 is that it does).

Click to expand...

Jono, it's not that the K10D through to the K20D, K200D, and K-m/K2000D don't produce DNG format raw files, they can; however they can't produce compressed DNG's as the K-7 and later cameras can which is the likely reason why Manu uses PEF for their smaller file size.

Regards, GordonBGood

Very nice overall. I've seen nothing but good from the K-5 so far.

However a couple of these pictures seem to be a little too red. Do you think it can be blamed on the lighting? I've taken the liberty of doing a slight correction; tell me what you think.

I now see why.

But I can't sell the 31 Limited to buy this!

Mark, your corrections are too green. Something in the middle might be fine.

Calibrating the screen might help..
--
DN