Saturday, May 12, 2007

Klaus Schroiff's K10D Review at PhotoZone

Klaus Schroiff, site owner and editor of the PhotoZone, recently has published his K10D full review and test report:-

I highly recommend to read this report of Klaus's (also for his other reports, without any hesitation) as his findings and conclusions are sensible and insightful, most importantly, comments and conclusions are usually made with good enough technical grounds and evidences. Nevertheless, his report is yet concise and easy to read. Of course, some of his particular conclusions and viewpoints made can still be subjective somehow but then this is of no difference from any other reviewers. At least, his tests and reviews are in a comparatively higher standard class of only few truly quality reviews belong to, amongst all other numerous sub-standard so-so "reviews" floating around the net.

Well, in Klaus's test report, he has actually reported something new, of which these new findings have never been reported by anyone so specifically before.

One of his new findings is about the obvious differences in resolution and the appearance of moires, etc. along the vertical versus the horizontal directions. With this strange and unique behaviour which I have reported here previously, Klaus did consider that it's simply a design flaw when he first saw the symptom. Later on Klaus was able to find out the possible cause to this problem, which is believed to be caused by the unequal strength of the Anti-Aliasing (AA) filter in the vertical and horizontal directions, more specifically: In the vertical direction, the AA filtering is much weaker than normal (i.e., very little dispersion to none, microscopically). And, in the horizontal direction, the AA filtering is yet stronger than typical (i.e., more dispersion).

Well, if we read into the details presented by Klaus. We can see that neither a too weak nor a too strong AA filter is good for the final image quality. Too weak a filter will create colour moires and very visible Bayer artifacts. Too strong a filter will decrease resolution significantly and blur the image too much.

It is found that Pentax takes the approach of using software (no matter it is in-camera or not) to do the demosaic job, as a remedial to the weak AA filter, vertically, which unfortunately just create even more blur and resulting in more loss in resolution. If 3rd party RAW convertors are used to convert the images, it can be seen that more serious and obvious colour moires will appear, although more resolution can be retrieved, also this will generate more noises too (see in the report for the noise level comparison by different convertors).

On the other hand, for the horizontal run, the image quality suffers from another extreme. So, this asymmetrical design of the AA filter, simply cause much undesirable effects which, however, cannot be handled by any RAW convertor with any software algorithm. In the end, one must see either more colour moires with better resolution (about vertically) or an image with less colour moires and noises but with inferior resolution and obvious blur.

The worst thing IMO is that the images will always look somehow different in vertical run against horizontally. I just wonder why Pentax need to design and make a camera in this way, of which this design and the produced results are indeed very strange and the image quality suffers and that no other DSLR makers else have ever done so, nor Pentax themselves for their previous models.

BTW, as we all know, the selling point of a 10MP+ DSLR is its higher resolution, however, the K10D by itself simply fails to deliver, no matter shooting in Jpeg or RAW, if only Pentax software is used. If one uses 3rd party RAW convertors, more moires and noises are resulted. Even worse, do note again that these are not symmetrical vertically and horzontally.

Whilst Klaus finally mentions this strange problem, which was actually first reported by him, is just a "Minor Con" in his final test report conclusion, I must regard this problem to be a *Major* Con, as this affects the image quality seriously and especially there is NO true solution to this killing problem.

Also, as for the "phantom" bright and dark pixels revealed under pixel peeping, it is actually a mystery to Klaus and all of us.

Finally, if you read back Phil Askey's K10D review, we can now all know why Phil Askey found the problem of "soft edged images" (in-camera Jpegs or RAW converted images via Pentax PhotoLab, same here for what Klaus has found), which led to a low rating in the image quality part of the DPR review. Well, now Klaus's latest findings and explanations have indeed given full answers to the questions raised by Phil Askey, of which there was no brief answer at all even by Phil Askey himself in his DPR K10D review. In fact, Phil Askey actually did not describe the problem as clear and as specific as what Klaus has achieved - Phil Askey just wondered and puzzled for the odds he saw and then he just (told us that he) "scratched his head hard". So, afterall, I think we must thank Klaus for letting us and all Pentaxians finally know the answers to this mystery! :-) My big smile to him and thumb up!


Olivier said...

Yadda, yadda, yadda ...

Anonymous said...

Regarding AA filter. Please don't make up facts without proof. As far as I know nobody took apart the filter "sandwich" to see the components. My GUESS is the AA filter in the one dimension is completely absent, not "microscopic" as you say.
Again you use poor reporting and exaggeration. This always leads to questioning your "facts". Sorry.

RiceHigh said...

I just wish to know what's the difference between your "guess" and Klaus's conclusion?

Anyway, if you just don't believe me and what I report, that's fine. What you need to do is to read the original review and test report. I do have provided the link, don't I?

BTW, below are two exact captures of what Klaus tells:-

"Now why's that ? The Pentax K10D has an extremely weak vertical AA filter. An AA filter is a low-pass filter that smooths the discrete "steps" caused by the sensor pixels. A weak or non-existent AA filter results in Moiree-effects - this is where the "false" colors come from."

"The AA filter aspect became pretty much obvious during the local lens tests where the vertical resolution turned out to be 50% (!) higher than on the horizontal axis - that's a quite unique behavior so far. All other cameras produced a symmetrical resolution characteristic. The horizontal AA filter is a bit more aggressive compared to the Nikon D200 or Sony Alpha."

Still any "poor reporting" and "exaggeration" this time??

Anonymous said...

YES..Still wrong and to quote him as you do "Admittedly I would have preferred a weak AA filter on both axis and not a non-existing + a strong one."
So what is it? weak or non-existant?
My POINT is neither you nor he knows definitively AFAIKT.... :)

RiceHigh said...

I really wonder what you're actually arguing for. The *key problem* found is about the unequal resolution measured vertically and horizontally, which the vertical resolution can be as much as 50% higher (but along with serious colour moires).

Klaus suspects that the root cause to the seen symptom is "an extremely weak vertical AA filter", which can be either "a weak or non-existent AA filter".

I really don't know why you keep arguing for if there is a vertical AA filter or not and try to prove that someone was wrong on this. But then in fact this is just a don't care case for the issue and the problem is still there, which is unarguable. Get it?

Anonymous said...

Obviously you don't "get it" Your pseudo-science and "tabloid" presentations are annoying and childish. My point was, and is, though the premise is the same, the AA filter is "different" in one direction than the other, you nor Klaus KNOW why except to speculate as to it being weak or not even present.
You state it has a weak AA filter but in effect you have NO CLUE if it is there or not. So this is just your "fantasy" and unworthy of being considered a "fact" or scientific" conclusion. It is just speculation, as was my "guess". Until someone removes the "filter sandwich" and dissects it, or there is confirmation from Pentax as to it's construction, then all you/we have is speculation. Speculation is fine and the conclusions are fine, just it is a matter of stating what is fact from speculation or what is fact from your wishful thinking that literally bugs me. Sorry, you have a place for comments and this is just mine. That is a fact :)

Anonymous said...

Anonymous, you are an asshole... treat "weak" as "weak or non-existent" and get lost from here

Anonymous said...

Ohhh the Ricehigh fanclub speaks.. w/ childish tongue. IQ of abot 30 I suspect.

Anonymous said...

"Now is this something bad ? Not necessarily. You may argue that this is a design decision because the potential resolution is roughly 10-15% higher compared to other Sony-based DSLRs (Nikon D200, Sony Alpha 100)"

"Finding a verdict for the image quality is tricky. The sensor has a very weak horizontal low-pass/AA filter. Consequently the potential resolution is very high - higher than on the Nikon D200 or Sony Alpha for instance - that is assuming you use a good RAW converter such as Adobe Camera RAW"

RiceHigh said...

But, there are some more words to follow:-

".. You may argue that this is a design decision because the potential resolution is roughly 10-15% higher compared to other Sony-based DSLRs (Nikon D200, Sony Alpha 100). That's assuming you use a capable RAW converter naturally - the Pentax PhotoLab or straight JPEGs do not qualify here. As we've seen the extra resolution doesn't come for free. This is probably also the reason why straight JPEGs as well as PhotoLab RAWs are comparatively soft because Pentax decided to apply an inferior software low-pass filter here resulting in sub-optimal results."

".. - that is assuming you use a good RAW converter such as Adobe Camera RAW, RawTherapee or CaptureOne. The Pentax PhotoLab as well as straight JPEGs deliver comparatively soft results. The very high resolution potential comes at some costs - artifacts. If you're into pixel peeping you may be able to find some horizontal jaggies, "false" colored lines and phantom pixels."

Well, I think the BOTTOM LINE for anyone is: Do compare the set of resolution test chart shots to see the differences and judge yourself.

Post a Comment

Related Posts

Creative Commons License
RiceHigh's Pentax Blog by RiceHigh is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.