[Class A]

Which shots are we talking about now? The test charts or others?

The outdoor shots. They seem too soft. I'll say that I'm not experienced with pixel-peeping at K-5 images (my K100D images look obviously crisper out of the box), but I'd be just a little surprised if that kind of softness were normal.

As far as my understanding goes, there is no way for sharpening algorithms to know deterministically know what transformation, if you want to look at it that way, caused the blur. Thus, the sharpening can't perfectly undo it, even if it can do a good job.

In general, you are right, but in the case of capture-sharpening a RAW file, the sharpener knows how the blur came about and can apply a deconvolution to exactly undo the blur. Different RAW converters will do different things, but in theory a complete reconstruction is possible.

Remember, anything that cannot be reconstructed from the blur is detail the sensor should not record anyhow, because it exceeds its resolution power.

Here's a link to a raw file from the K-5:

Thanks. I played with it a little. See below a quick n' dirty attempt to sharpen the original K-5 file so that it competes with the K-5 IIs file.

The order of the images is:

1. K-5 as sharpened by you.
2. K-5 as sharpened with LR.
3. K-5 IIs (called "RAW" but apparently already sharpened).

Note that what you labelled as "K-5 IIs RAW" (in your "best attempt") shows faint sharpening halos around the tree tops. Maybe your RAW converter has some default-sharpening which isn't exactly zero. Maybe they are JPG conversion artefacts (or the result of "sharpening for screen").

I had to muck around a bit with exposure and colours to make the K-5 shot resemble the K-5 IIs shot. Your RAW converter camera profile (or whatever produced the "K-5 IIs RAW" shot) seems to be quite different to LR when using the "embedded" camera profile.

I'll be the first to admit that my LR sharpened version could be called oversharpened but it would print just fine. I mainly wanted to show that one can extract apparently lost detail with some aggressive sharpening.

Finally, I'm not a 100% convinced that the K-5 shot has received optimal focusing. The unsharpened version (see your "best attempt") is very fuzzy. The sharpened version exhibits almost no sharpening halos. At 1:1 level, I'd expect to see some minimal sharpening artefacts. In other words, it hasn't been sharpened enough, AFAIC.

Finally, the K-5 IIs shots have higher mid-tone contrast than the K-5 shots in your "best attempt". This has nothing to do with AA-fitlers or sharpening. Whatever you or the RAW converter is doing, it does not create a plain level field for both cameras.

[Class A]

... Fuji, the way they fixed the issue with their newest camera (by having a "random" arrangement of RBG pixels on the sensor, similar to how color film was made, has eliminated the need for an AA filter AND produces razor-sharp images with no moire!

• The Bayer-pattern used buy Fuji is not random. It is just larger than the regular 2x2 pattern.
• The pattern does not eliminate moiré. It makes it less visible, but trades this off with a higher chance of ocurring.
• The approach does not create "razor sharp" images comnparable to AA-filterless images, since it trades-off resolution in return for making moiré appear less visible.

BTW, Adobe is still not supporting the Fuji Bayer-pattern with their RAW converters. A completely random pattern would be next to unsupportable.

[frogfoot]

User reports suggest that it has been fixed

Great to hear it

[Tord]

Please remember when forming opinions of value that there is a whole world outside the USA. The UK list price for the IIs is £150 more than the II. That's about $240 or, to put it another way, not a bargain.

For better resolution $240 seems cheap to me, but to upgrade from a K-5 to a K-5 IIs is a steep hill to climb! I jumped ship, sold the K-5, and bought a D600 at a fair price, I thought (a little over 1660 pounds - 18,000 SEK).

[rparmar]

I agree that a contemporary lens should be used for such a test. Older lenses do not have the same micro-contrast, which is particularly important in such an evaluation. The DA35 Limited is a good choice, perhaps.

From the responses to the level-headed review, I can only say that people really need to stop hyping every new product as the second coming. It was quite obvious from other examples (Nikon) that this model would be a minor incremental improvement over its predecessor and nothing more.

Nonetheless, I for one applaud Pentax for giving us the option of no AA filter. Far too expensive in the UK however.

[baro-nite]

Here's my attempt from the K-5 DNG kindly provided by Adam:



And again Adam's jpeg comparison:



I used R-L Deconvolution in Raw Developer (Iridient), radius 1.0 pixel, 50 iterations. Thanks to Class A for clueing me in that R-L Deconvolution is, theoretically at least, the correct method to remove the blur caused by the "AA" filter. I don't know if these settings are correct for that purpose; I picked 1.0 pixel radius based on my very limited understanding of how these filters work, and 50 iterations because it is the maximum available in Raw Developer.

[baro-nite]

Hmm, I should have looked beyond 100% before posting. Clearly there is noise introduced by RL deconvolution at 1.0 pixel radius. Simply going by eye, somewhere around .40 to .42 is where noise starts to kick in. Here's 0.42:

File Date

View Picture EXIF

Very similar to Adam's mildly-sharpened version of the K-5 shot.

Note that what you labelled as "K-5 IIs RAW" (in your "best attempt") shows faint sharpening halos around the tree tops. Maybe your RAW converter has some default-sharpening which isn't exactly zero. Maybe they are JPG conversion artefacts (or the result of "sharpening for screen").

Could this simply be aliasing caused by the filterless sensor? Edit: Or is that always going to be false color?

---

Last edited by baro-nite; 10-30-2012 at 02:38 PM. Reason: addendum

[nanok]

The outdoor shots. They seem too soft. I'll say that I'm not experienced with pixel-peeping at K-5 images (my K100D images look obviously crisper out of the box), but I'd be just a little surprised if that kind of softness were normal.

no, it is not.

thanks for the demonstration

for Adam: it is quite clear to me classA was not out to prove anything at all, he's just voicing valid concerns, and making sure we're comparing apples to apples, and staying objective. unfortunately "it's obvious IIs has better resolution, why don't people get it" is not what being objective means. the problem here is that it's easy to assume the k5IIs will be sharper, and find the proof when we're looking for it.

we need a physicist to outline the rigorously scientific test procedure to follow, in order to clarify what's what: more resolution? just less sharpening needed? etc . where's falco?

[JeffAHayes]

• The Bayer-pattern used buy Fuji is not random. It is just larger than the regular 2x2 pattern.
• The pattern does not eliminate moiré. It makes it less visible, but trades this off with a higher chance of ocurring.
• The approach does not create "razor sharp" images comnparable to AA-filterless images, since it trades-off resolution in return for making moiré appear less visible.

BTW, Adobe is still not supporting the Fuji Bayer-pattern with their RAW converters. A completely random pattern would be next to unsupportable.

Thanks for clearing that up for me... I've been wondering if I got a true story on this camera, or not, when Britain's Digital Photo magazine recently did like a 10-page "review" on it (I was getting the mag. at Costco here in America, but it's not there any more, and so far my email to the mag. has remained unanswered).

At any rate, the "review" was pretty obviously at least as much ADVERTORIAL as it was "review." Every page even said, at the top, "in association with Fuji" (or something similar -- don't have that issue handy at the moment). At any rate, they used a professional model for the shots for the review, and in one case showed a full-sized crop of just her eye from a photo above, clear as a bell, claiming that was due to Fuji's new "random pattern" that eliminated the need for the AA filter. Since that's the only place I'd seen this, it was a bit suspect to me, so I figured I'd throw it out here and see what I got back.

That was one of two British photo mags I could buy at Costco here in S.C. Digital Photographer disappeared this past summer; Digital Photo just in the past month or two. My last email to Digital Photographer was returned as being an invalid email address, so I fear that magazine has ceased publication. Shame. Each issue, while expensive, was more like a coffee-table book than a magazine. Both beat Popular Photography, in my opinion, although that's the only one I subscribe to, since it's cheap, lol (it was roughly $100/year to subscribe to either of those British mags, sometimes more). Last couple issues, both had reduced American subscription prices, though. I think they were getting desperate.

When you consider that NEWSWEEK will cease print publication in another two months, well, I'm guessing we'll be missing A LOT of mags we used to see within a year or two (unless we get them online).
Jeff

[Winder]

• The Bayer-pattern used buy Fuji is not random. It is just larger than the regular 2x2 pattern.
• The pattern does not eliminate moiré. It makes it less visible, but trades this off with a higher chance of ocurring.
• The approach does not create "razor sharp" images comnparable to AA-filterless images, since it trades-off resolution in return for making moiré appear less visible.

BTW, Adobe is still not supporting the Fuji Bayer-pattern with their RAW converters. A completely random pattern would be next to unsupportable.

The Fuji XP1 does have "razor sharp" resolution. Look at the resolution tests for the Fuji 35mm lens.
Fuji is working with both Apple and Adobe to incorporate RAW support, and SilkyPix does support Fuji RAW which is shipped with the camera. I'm not sure why you think it is "unsupportable".

[Class A]

I'm not sure why you think it is "unsupportable".

I wrote "a completely random pattern" would be next to unsupportable. The Fuji pattern is not random.

[Winder]

I wrote "a completely random pattern" would be next to unsupportable. The Fuji pattern is not random.

OK. I misunderstood you comment.

[Class A]

Perhaps moire is at play, assuming it influences the raw data that the sensor captures before the image is processed.

Seems unlikely. My bet is on the new JPG engine and I hope you don't see the same mush in a K-5 IIs RAW file.

Are you still working on a more comprehensive K-5 vs K-5 IIs comparison? In addition to validating the AF microdadjustment settings on the cameras for the FA* 85/1.4, you should probably also try a different lens model (e.g., a macro) and preferably use the same lens copy on both cameras.

On a somewhat related note it would be great to learn more about the f/2.8 sensor of the K-5 II(s). Are you planning on doing any comparisons in the future?

[Adam]

Seems unlikely. My bet is on the new JPG engine and I hope you don't see the same mush in a K-5 IIs RAW file. Are you still working on a more comprehensive K-5 vs K-5 IIs comparison? In addition to validating the AF microdadjustment settings on the cameras for the FA* 85/1.4, you should probably also try a different lens model (e.g., a macro) and preferably use the same lens copy on both cameras. On a somewhat related note it would be great to learn more about the f/2.8 sensor of the K-5 II(s). Are you planning on doing any comparisons in the future?

The same lens was used for the test, but we can try a different lens (say a 77mm or a D-FA 100mm) if it'll make you feel more comfortable with the results.

Regarding the F2.8 AF thing, I don't think we're going to go into that much depth. I don't think any of the conclusions made in the review are going to be changed based on such investigations, though I'd be happy to take a look at them if a user decided to do a test. From what I've been reading on the forum in the past couple of days, it seems that users are trying very hard to find changes and improvements in the K-5 II beyond those that have already been listed and/or discovered; to me, this is concrete evidence of the fact that loyal Pentax users aren't going to be satisfied by this release. I'm keeping my fingers crossed that the new Pentax APS-C camera, if one should be released anytime soon, will be a "direct upgrade" over the K-30, and not something that makes compromises in areas such as video and processing speed.

[CtrlAltDel_ru]

I'm keeping my fingers crossed that the new Pentax APS-C camera, if one should be released anytime soon, will be a "direct upgrade" over the K-30, and not something that makes compromises in areas such as video and processing speed.

You several times mentioned about new Pentax APS-C camera. Do you have some insider info?