This article presents the text of a recent comparison test I carried out over on the UK PentaxUser forum. This was a forum thread with plenty of comment in between my ongoing testing results, but here you can see the description of the full test uninterrupted.
The shots were offered up for use on the DPReview website, full DNGs for download and study purposes. As they are not my shots I won't be reposting them, so you'll need to check them out over at DPReview ....
Studio shot comparison: Digital Photography Review
So.... here's how I see this issue, as much as I can place it into some logical context :
We have two choices, which provide a different initial input at capture - one (K5II) deliberately blurs the image by means of an AA filter, the other (K5IIs) does not.
With a blurred input (K5II) a compensatory action is applied in processing - this could be in-camera processing or a default deblur algorithm applied in a RAW converter program through the default camera profile for that program. (Then the user has the option to fine tune this in the software). This process is known in workflow terms as 'Capture Sharpening'. So for the K5IIs you can eliminate this sharpening stage.
How significant is eliminating capture sharpening? As far as the actual work involved is concerned, not at all. The default profile for RAW conversion can simply remove it, or the user can simply reduce the amount applied directly. We should get basically the same initial capture result with no effort, so no gain there.
The important question is how this small difference in reaching a default capture sharpening state will affect the potential of the image to absorb additional defining, detailing or sharpening processes later on in the workflow - call it 'output sharpening'. This could have added value in two ways - 1) to produce a final output with detail and clarity qualities that exceed that (at any given size) of an equivalent image shot with the standard K5II and DeBlurred with Capture Sharpening, or 2) to produce a larger output (say a print) without any further loss of quality beyond the maximum achievable from the equivalent image shot with the K5II.
If we can answer that question then we have a basis to decide on the value for money equation, and / or the 'risk of moire' equation, all depending on our own personal output needs, or desires!
Considering all the above, looking at this up front, I'm going to throw caution to the wind and claim that in my view, unless you are going to be pushing the boundaries of output (my No1 & 2 above), then the differences at capture stage between the two (allowing for the default compensating DeBlur algorithms applied at Capture Sharpening stage to the AA filtered image) will not be significant to the majority of users.
As far as the second part goes - what is the potential for increased clarity / size and quality at the output stage - then we have to run shots through our own individual workflow, using our standard tools, and attempt to compare the results with identical settings, and then push the boundaries to see if you can get further with the K5IIs files without quality loss.
The DPreview K5II & K5IIs Test Shots 1 - First Views & Capture Sharpening Stage
I'm looking at the shot that includes a lot of bright and detailed items, like the Martini and Baileys bottles, Micky Mouse, Queen of Hearts card, blue watch etc etc. So, to follow my methodology I first need to create a level playing field to eliminate any differences created by default camera settings or different RAW conversion software acting on any of these settings. Then I can render the files using the exact same process for both - this gets me to the 'Capture Sharpening stage', but with any processing variables eliminated.
The DNG & Metadata Problem
These files are DNGs. This format can record a great deal of exif and camera propriatory metadata, and Adobe software can read most of this data. Where the fields for this data match up with Adobes own exif or other settings fields there is the potential for the Adobe Camera Raw plugin to act on this data upon opening the file and add these adjustment actions to its default profile when the file is opened.
The problem when making comparisons or considering what images look like to different people is that we don't have a way of knowing exactly whether these extra processing settings are being applied or not. This is further complicated if people are using different RAW converters. So the level playing field can be hard to establish. I will deal with this issue by using Pentax Digital Camera Utility (which can read all the camera-set settings & adjust them), leveling them, then rendering the files both to JPEGs and then 16bit TIFF with the same sharpening settings. They can then be examined closely in viewing software.
Camera Generated JPEG Previews
I have now opened both files in PDCU, displaying their preview windows side by side. This window shows the image on the left, and on the top right all the camera shooting metadata with a histogram at the bottom. The image I see here has not been rendered on my computer - it is a small jpeg rendered by the camera at point of capture, processed with all the camera-set parameters applied, then embeded in the DNG file. This is the same image that is displayed on the camera LCD screen just after shooting. As it is edited to reflect all the settings applied on the camera (eg shadow / highlight corrections, contrast, sharpness & fine sharpness) it is important if making comparisons on the camera (like in a shop), or intending on shooting JPEGs, to be absolutely sure all these settings are equal over both cameras. This embedded JPEGs are also important here as they represent (in compressed form) each individual cameras JPEG output.
The Metadata
Here is the metadata that is relevant to this testing. If I do not list any particular parameter, then it is set to its O position :
The following data is recorded identically in both shots : Both cameras are on firmware version 1.00
Lens - D FA Macro 50mmf2.8
Shutter - 1/20 sec
Aperture - f8.0
Mode - Manual Exposure
White Balance - Manual
ISO - 100
Colourspace - sRGB
Custom Preset - Bright
Contrast - +1
Sharpness - +1 (Fine Sharpness is 0)
Drive Mode - Self Timer 2sec
Flash - Off
Shake Reduction - Off
The following parameters are recorded with different settings between the two :
White balance fine tune on the K5II is set to Green +1, but on the 's' it is 0.
Metering mode on the K5II is Centre-Weighted, on the 's' it is Matrix Mode.
Finally, 'Spurious Colour Signal Reduction' was activated with a value of 80, the same for both shots.
There is a noticeable green cast across neutral tones on the K5II image, so I assume the WB fine tune setting was a mistake, or not noticed. Possibly the metering mode difference also was not noticed, but there is no significant effect on the exposure between the two.
Visual Impressions
The test images are well focused, bright and colourful. In the PDCU preview window I am able to go fullscreen and zoom to 100%, which I did do and examined numerous objects comparitively. There is a small amount of depth of field variation within the scene between the items. This has the effect when studying the shots up close of making some items appear fractionally more defined than in the opposing shot, but there is no obvious pattern to how these DOF variations have recorded between the two. But manual focussing was used, which could explain it.
The histograms are virtually identical, with a tiny amount of black clipping and the white point set at about 93% across the scale. The 's' image gives the impression of being fractionally higher in contrast, but this is not supported by the histogram.
Colours on the K5II image are fractionally more vibrant, which is interesting considering the comment about contrast on the 's' I made just above.
It should be clear by now that there's really not much between them by any measure. Of course, we are not talking about rendered RAW files yet - these are the camera processed JPEG previews. As far as these JPEGs are concerned, I cannot see any consistent difference in 'sharpness' between the two.
Preparing for Rendering
Now I will open each file in PDCU's main editing window. This changes the operations - now PDCU carries out 'on the fly' rendering as adjustments are made, although it only generates a preview with low resolution. This allows adjustments to be displayed accurately as you go, but PDCU can only operate on one file at once, so side by side comparisons will not now be possible until final rendering at full resolution and the new JPEGs or TIFFs are opened in another programme.
I opened both files one by one, and restored the contrast and sharpness settings to 0, and ran through all the other settings to be sure they were the same, including the WB fine tune. I deactivated 'Spurious Color Signal Reduction' as it seemed to have no effect - but who knows what it would do if the file was opened in ACR or something else? This all gives me confidence that there will be no differing processing affecting the shots caused by any unknown matadata field. It would be more difficult to have this confidence in another RAW converter, as the metadata fields in the file may not all match up to the programme parameters.
I then examined the shots closely again, although only individually.
First The K5IIs image shows the same slight variation in DOF throughout scene. There is nothing new to report in this respect. There is one object, a white rectangular thing at the left base of the Baileys bottle. This has a fine ribbed texture.
The surface of this object is pretty much wiped out with arrow shaped green bands - I can only assume this is a moire defect. Then the K5II shot. I of course immediately zoomed in to the 'moired' object, and , guess what, there is no arrow shaped green banding on the surface. So there's a clear difference with that. Scanning around I cannot see anything further to report - both images appear the same at this stage as their JPEG previews previously.
I saved out (rendered) 3 versions of each - 16it TIFF, '4-star' JPEG and '2-star' JPEG.
The Rendered Files - Capture Sharpening Stage Comparisons
Please bear in mind that these initial comparisons are on equal, processing neutralized images at the first stages - this is the 'Capture Sharpening' stage. This does not yet answer what the final output potentials are of either file.
However, I believe that these initial comparisons to be accurately representative of the JPEG output from these cameras. This is based on the fact that the rendering and metadata were neutralized in PDCU, and the output shows no significant difference to what the first camera-generated embedded JPEGs looked like.
I now have 6 files to look at. First I chose the maximum quality JPEGs and opened them in my favourite image viewer - Fullscreen Image Viewer. This displayed them fullscreen and both shots looks beautifully presented, clear and detailed. There was no discernable difference (apart from the moire) when flicking between them at fullscreen size.
Then in my imatch program I opened them in the 'side by side' window. Here I have full zoom and scrolling control and both images move together, so I am looking at exactly the same items at the same time at the same size. Also there is a 'loupe' tool for instant 100% zooms on a smal spot.
I explored the images extensively using these viewing methods. I also opened the 16bit TIFFs and carried out the same exploration.
I could sense at times some slight difference at certain spots, but there was no conistency to this. In total, on balance, I can honestly report that I was unable to identify any discernable difference between these files, either in their TIFF or JPEG versions.
For that reason I believe I can fairly argue at this point in the process, that if you only wish to shoot 'in-camera' JPEGs, then there is no benefit whatsover to the K5IIs model.
So thats that! The fun will really begin when we start to explore the effect of the next stage in the processing chain - output sharpening.
Checking the File Data - A Shock Discovery
We would all expect, logically, that the 's' version of this shot would show more definition (less blur) 'out the can'. This should be inherent by design. There is no AA filter over the sensor of the 's' camera creating deliberate blur. The files I have created so far represent an equalising of this 'out the can' stage - DNGs with any variables nuetralised through identical RAW conversion, with a default level of Capture Sharpening - or do they ??
Here is the file data. Both shots have precisely the same figures :
4928x3264, 300x300dpi, Bit Depth 48, Uncompressed, File Size 92.1mb
This info is contained with the Windows file properties box. But suddenly, I was shocked to find a real surprise - there is some 'advanced' metadata recorded by Windows further down. I was just routinely scanning when suddenly I noticed a discrepancy - there are two fields, 'Contrast' & 'Sharpness'. On the K5IIs image these two fields show 'Normal'. But on the K5II image these fields show 'Hard'.
You can change this data with three options - 'Normal, Soft, Hard'. But considering the meticulous neutralizing process I went through, how could these fields have recorded differently when all settings on these files were identical?
I straight away checked the JPEG versions also. And again, exactly the same - K5II version records 'Hard' and K5IIs 'Normal'.
This raises the obvious immediate question - is there some mystrious, hidden metadata value buried deep within these files, and being acted upon by PDCU to affect the default opening sharpening state, a hidden process to us, not controllable by the user, not seen in the program, unrelated to the user adjustable parameters, but picked up and mapped to the Windows standard properties fields?
I am aware that individual camera models have specific default profiles for RAW development - is it possible that the standard K5II model has a higher amount of default contrast and sharpening built into its default profile, in order to compensate for the blurring effect of the AA filter - and that the 's' model has a significantly lower amount of this contrast and sharpening?
Could this in fact be the reason that both the camera-generated JPEG preview, and the the PDCU rendered files, do not actually display any obvious differences in sharpness or resolution?
All of this mysterious default profile stuff is well beyond my knowledge, so I appeal here for any other views and experience of this.
Re-Running the DNGs through ACR
So, to further confirm my suspicion I'm running the two DNGs through Adobe Camera RAW and will use PS Elements to save out new TIFFs and check if the same metadata fields have recorded. This would confirm that there are differences buried inside the default profiles. I'll list the camera RAW opening defaults :
Exp / Recovery / Fill All 0
Blacks 5
Brightness +50
Contrast +25
Clarity / Vibrance / Saturation - All 0
All Sharpening & Noise Reduction settings are 0
There are differences in the Temp & Tint settings : K5II - Temp - 4450 / Tint - +32. K5IIs - Temp 4200 / Tint +31. Both shots were taken with a manual white balance reading, but clearly both were not exactly the same. This difference also confirms that this camera recorded data (WB) is being read and acted on in this different RAW converter.
However, the sharpness setting is 0 in ACR, but in the files the camera settings were +1 for sharpness, so ACR has not acted on this user set parameter. This shows how difficult it is to judge how the same file will be dealt with on default opening in different RAW converters. My guess is that ACR does not map the camera custom image settings sharpness field to it own equivilent parameter.
I launched the files on conversion into PS Elements and saved out TIFFs. These files show some dimension differences (4928x3264 - 240dpi) & the file size is similar at 92mb. But, this time, strangely, both the Windows Properties 'Contrast' and 'Sharpness' fields show 'Hard' for both files. Not very helpful, as it doesn't really confirm my theory. We may never know exactly how these fields are applied in the different programs nor how they are copied over to Windows Properties ??
I saved out full quality JPEGs from PS Elements also and did the close up comparisons. This time I am more confident that with the 100% loupe tool I was seeing, with consistency, a very small, but just noticable, finer detailing in objects in the K5IIs shot.
Could this be caused by the ACR converter not applying additional sharpening to the K5II that was applied by PDCU - could this be due the two programs reading and interpreting default profile data buried deeply somewhere in the files?
But, this doesn't change the conclusions for now - people shooting JPEGs are not then going to run their shots through ACR, I would presume, so as from the first testing stage I can still say that for that use solely, with no further processing planned, there is no apparent advantage to the 's' version.
So far, with default RAW conversion - Comparing the K5IIs outputs side by side (ie one from PDCU & one from ACR) showed no noticable differences in definition.
Comparing the standard K5II shots side by side (ie one from PDCU & one from ACR) shows consistently a noticeable higher definition to the image processed in PDCU. This supports my theory that it is only PDCU that is reading hidden metadata and applying additional compensatory sharpening to the standard model shots. DeBlurring Methods & Comparisons - through to Output Sharpening
The idea behing all this is that it could form the basis for informed decision making based on an individuals output needs. For this reason it is now time to do the heavy lifting - ie put the test images through a normal workflow and compare the results. Also I will try and achieve an equal appearance of definition and clarity in each shot in order to compare side by side for quality differences.
First I will run through a typical workflow for myself. I will not be adjusting colours, but as well as definition and detailing I may have to tweak contrast and brightness to extract the best I can out of each file.
A Typical Workflow for me
First I'm going back to PDCU. Capture Sharpening does not have to be a default or minimal amount - it makes sense to add a certain amount of extra sharpening at RAW conversion point as this is point where you may have your best stab at extracting all the quality you can from your image - you're at full bit depth, and working non-destructively. So this is the time to get a good sharp starting file, but leaving latitute to work selectively and more intensly on certain details later (output sharpening).
I start by opening the K5II image in PDCU, and I make the following adjustments to improve the image, considering the aim to output to a high quality and / or big enlargement at a later stage :
Exposure Value + 0.5 stop
Highlight Adjustment DR Expansion + 0.5 unit (to recover some areas of clipped white resulting from the EV increase)
Contrast +1
Low Tone Contrast -3 (Pulling black down)
Sharpness +3
The sharpness setting is not something that can be easily evaluated in PDCU - it has to be treated as a kind of default based on the type of scene and experience with it. In this case its a crisp, sharp shot of hard objects, so I think + 3 will be about right.
I save out a TIFF and open it in my viewer. Now we're cooking! This version has a real sizzle, a presense that was lacking in the default setting versions up to now.
I repeat the rendering with the exact same settings on the K5IIs file.
First Comparisons following RAW conversion as above
The level of clarity at 100% loupe view on both shots is a bit better, not much, but a clearly noticable amount. These versions could both be printed at average sizes and look fine at any normal distance.
Now I can say for sure that there is a more consistent, slight, higher level of definition on objects in the K5IIs shot - very slight, but it is there on more objects. Now I can say for sure that the inherant extra clarity at capture on the K5II has been revealed more clearly through a combination of better editing (the contrast and exposure enhancements) plus the additional level of sharpening applied to this output.
Now I want to see if the K5II shot can be made equal through the same processes simply by extra sharpning. I re-run the rendering, this time with +4 (the max) for sharpness, save out and compare side by side with the earlier K5IIs shot (which had +3 sharpness).
I have evened the gap now for sure - there remains a touch of extra bite to objects on the K5IIs image (with +3 sharpness) than those on the K5II shot (+4 sharpness) - a very tiny difference. There are no apparent negative effects on the K5II shot due to the extra sharpness applied.
An Ongoing Workflow through to Output Sharpening
I am now sticking with the last two images compared in 16bit TIFF format.
Now I' m going to try some more agressive actions, aiming to enhance and add clarity to the image, and bring out visible details - with the aim being to prepare these shots for quality printing at a very big size.
I'm not going to make any further contrast, brightness or colour changes, and I'm not going to use any noise reduction - lets just stick to definition, clarity and detail enhancements.
For now I'm going to continue using my own chosen software for this work - later I will try and do similar things using standard Photoshop tools.
A) Definition (micro contrast) in Sagelight 4.0 with identical settings
First up is the K5II shot, into Sagelight 4.0. I have two 'Definition 'sliders', Strength & Radius. Default radius is 20 with the maximum 300. The' help' says that low values 'are more like standard sharpening', so I'll keep 20 set. I zoom to 100% and try a few strength settings between up to 100. The affect on the image is is very positive - a slight extra punch to sharpness, plus a tiny boost to contrast at a quite local level. Globally, viewed at fullscreen, this removes a slight hazy hint, and I now have a very deep and sharp looking picture. I settle on the strength of 50.
Next up into Sagelight is the K5IIs TIFF - I make the same definition settings and save out again. Just for reference, the un-compressed TIFFS saved in Sagelight are remaining at 92mb file size. I am saving out full-size JPEGs also as a record of each stage - these are coming out at 17mb.
So, now, side-by-side 100% comparisons show a clear result......the K5IIs image is pulling ahead in clarity terms. I can more easily and consistently clearly see better definition in the 's' image. It seems as though the additional and more aggressive 'definition' processing is pulling out something more from the 's' shot than the standard one.
There is still nothing in it though at a full-screen view. The differences start to creep in at about the 70% magnification level.
I'm starting to take a good look at shadow areas now for noise levels, and checking on the clear textured surfaces to look for any disruption to patterns or smoothness. I can report now that at this stage in the processing chain I am not seeing any sort of issues or quality degradation in the shadows or other smooth textures areas, on either of the images.
B) Adding further high detail with Topaz Detail PS Plugin
The images I've got so far look very crisp on screen and have a nice 'presence', and they will certainly look very good to average printed sizes. But I know from experience of preparing my shots using the same workflow steps, that if I got them printed to very large (possibly A3 or bigger), I would not see the same crisp 'presense' in the print as on screen - there would be flatter areas and larger areas of the same surface would take on a slight OOF effect. To correct this for these big prints I need extract even more edge definition at both the medium and fine radius levels.
Topaz Detail plugin provides an easy way to achieve this, with both 'boost' and 'strength' sliders in three groups - working on 'large', 'medium' and 'small' detail levels. This operation will produce an exaggerated effect for screen and web use, but should be just right for really big printed output - such as my favourite 70cm / 100cm poster prints.
First up is the K5II shot. For those who know the plugin, you can do some serious damage with it, but I'm taking a very subtle approach. After some experimenting I settle on the quite minute settings of :
Small Detail - 0.08
Small Boost - 0.06
Medium - 0.05
Boost - 0.04
Blur Size - 0.03
Saved out to a new version of the TIFF and viewed fullscreen this image is really sizzling now! The surface of the objects has a presence that almost looks like you're looking at the real object. But I think I've reached the limits to which I can push this process. On the smooth surfaces there is now a boosted texture that pushed any further might start to look exaggerated.
I carry out the same operation on the K5IIs shot and open them up side by side.
This time I am not seeing any further separation between the two. The very small difference previously noticed at the last stage has been maintained. There is not enough exaggerated texture to cause problems on very large prints I think. So I would be happy to send either of these to the printer for very large output. I am confident that the K5IIs version has a small amount of added visible detail. C) Trying to extract even more from the K5II image to equalize it
But, I will carry out one more test in the workflow - lets see if I can push the K5II image just a bit further in Topaz to check if I can equalise the detail without overcooking any textural roughness. I'm reloading the original version from the stage before Topaz, just after the Sagelight definition was added, and I'm going to apply higher values and see how far it can go.
I start out by doubling the values from the sliders set before, but this casues highlight microcontrast to look 'blown', starting to look speckly. So I step back and settle on :
Small Detail - 0.10
Small Boost - 0.10
Medium - 0.10
Boost - 0.08
Blur Size - 0.03 (unchanged)
Opening the result in Fullscreen Photo Viewer, I am just astonished at the level of detail now being seen. The objects have a more solid feel with textures that I imagine I could sense through feel. There is slightly less deep tone contrast, but I suppose I could replace this if I wanted to.
First I compare at fullscreen the previous version of this shot (with the original Topaz settings) against the new one. I would proably now choose the earlier version for screen use now, and the latest (with the extra Topaz work) for the printer.
Now the real test. How does the earlier version of the K5IIs image stack up against the latest K5II image with the extra enhancements?
This time I go for the side by side magnified views again. I can clearly see now, on the highly detailed and intricate objects, that there is the same level of clarity, & edge crispness on both shots. But, in the K5II image, on smoother textures and in shadow areas, I now see some negative affects of this processing. Smooth areas have an added texture, a kind of roughness, that is exaggerated, and shadow areas have lost contrast, appearing a bit hazy.
It seems that this intensive detailed work is starting to break up the integrity of the image data, at least in smoother areas. I do not see the these negative effects on the same parts of the K5IIs shot, so this appears able to absorb more and hold up under this intense processing. If I had to choose at this point and send one for printing at a large size , it would now 100% be the 's' version. This is the first point in my testing that I have reached a clear delineation mark in this decision. Proving The Theory
This is the theory that has developed so far - I accept that there is natural, inherent added sharpness and detail to the K5IIs images at capture point, by design. But I have seen that right at the beginning of the processing chain (camera JPEGs, or basic RAW conversion with simple 'capture sharpening') that the blurred input of the K5II is easily, in some cases automatically, corrected with the result that the images are made equal.
I have also seen how the shots respond to my own personal workflow for highly detailed images intended for large scale printing. The K5II image has reached a quality limit that makes the choice clear - I would need to use the K5IIs shot at my intended print size.
It seems reasonable then to assume that at lower levels of detailing in processing that images from both cameras will simply 'sharpen up' at a fairly level rate - that at any given point you could just stop, crank up the K5II amounts a bit and equalise the shots.
I want to confirm the theory with a few quick tests at the lower levels, trying out some other common techniques.
D) A few Comparisons with simple Photoshop Sharpening
So, back to Adobe Camera Raw. Previously with this converter I was just checking on its output with default settings, and I discovered that it doesn't read the metadata in the same way as PDCU. This time I am aiming for a good looking photo, a proper workflow, so I'll make some basic adjustments to each shot.
First up is the K5II DNG. I settle on the following ACR settings :
Exposure - +0.10
Recovery - 15
Fill light - 0
Blacks - 5 (I'm keeping the original slight black clipping)
Brightness - +70
Contrast - +33
Clarity - +40
Sharpness - 50 with Radius 0.7
Converted into PS Elements I save it as a 16bit .psd file. I carry out the same ACR conversion with identical settings on the K5IIs DNG and open both images side by side in Elements.
With the K5II image active I now open the Sharpness Dialogue - maintaining the radius of 0.7 (carrying on from the ACR setting), 'more refined' ticked (I mean, I want to be more refined, right?) and Preview ticked, I can quickly see the effect of my percentage movements on detail in the K5II image at 104% magnification - the K5IIs image sits right next to it with the exact same view.
The figure of 23% produces an identical result on both shots. K5II 23% = K5IIs 0%
I try out the same setting on the K5IIs shot, and it jumps ahead again, maintaining the differential. I crank it up to 40% and the fine edge sharpness firms up noticeably.
Back on the K5II image I use 'Undo' and return it to its opening state. I now try to match it up again with the 's' shot beside it.
This time the K5II shot equalises at 60% - K5II 60% = K5IIs 40%
I increase the value on the K5IIs shot to 80%. This looks good, but seems to be about the limit of what I can expect to get out of this particular sharpening tool. Much further and I'm seeing roughness appear on the smooth textures, and fine edges (like the edges on the gold lettering 'Baileys') are starting to become slightly jagged. Still at 104 % view though. Further out and the image looks good.
This time the numbers start to race away - K5II 140% = K5IIs 80%.
I go for a 154% view - this gives me that blue watch in the lower right corner filling the whole image window, looking at them side by side. I am surprised to find that, once again, there's nothing in it! The K5II image has held up, with no visible artifacts reducing its quality, at the 140% setting.
Ok, one more - I've already got this image as far as I feel it can go with this fairly simple sharpening process. Lets crank it up one more time. I 'Undo' both images to keep this non-destructive.
This time I find I can take the K5IIs shot to 200%, and its still holding up. There's only a tiny increase in edge 'jaggedness', and a little more surface disruption, not much, only really visable over 100% view level.
But now I'm really stretching the K5II image. To match up and equalise the amount of edge detail and clarity on fine objects I have to take it to 300%!
K5II 300% = K5IIs 200%. This time I have reached the limit again. The 's' image is acceptable - an increasing presense of surface roughness is creeping in and further signs of jagged lines. But the K5IIs image has gone beyond what I would accept - certainly for any large printing. I see higher levels of surface textures breaking down. For example on the face of the woman there is now much higher graininess on the K5II shot - the 's' shot has smoother tone here, and this pattern repeats across most of the single texture areas, like the colour charts for example.
How will this now affect the choice? Now I need to compare my two efforts to see which would be my first choice for my large print in terms of 'good' detail.
I go back to my earlier version of the K5IIs image (run through the PDCU - Sagelight - Topaz workflow) and I see again the excellent 'sizzle' & 'presense' now apparent, with little degradation on smooth tones or contrast breakdown to detract when viewed big.
Side by side I check this image against my PS Elements sharpened version (200% value). At 100% view I immediately notice that the very fine small details are sharper in the Photoshop version, although with noticable jagged lines. The 'Topaz' version is less crisp at this fine level, without the jaggedness, but also includes added boost to larger areas and textures. Looked at fullsize the Photoshop version loses it's impact - it has a smoother and flatter appearance, slightly dull. The 'Topaz' version has better separation between the items, and that more real feeling to it, like you can touch the objects.
I realize this must be because my Photoshop sharpening was limited to a single radius level, 0.7, so just the finer details. The Topaz work has acted on a wider range of detail automatically, and the benefits are clear to see.
Remember I had reached the upper limit of sharpening in PS Elements on the 's' shot and exceeded that in the K5II version (with exaggerated jagged fine lines and rough textures appearing) - so my choice is clear. For this particular style of highly intricate and detailed subject, both for screen viewing and large printed output, I clearly want to use the K5IIs image - this has given me undoubted visible benefits. CONCLUSIONS - MY 'NEW TAKE' & A PERSONAL CHOICE
OK - its taken a while, and just on a single sample scene! I don't usually bother getting so close up & personal to the image. But I have enjoyed the effort and have learned more about how shots respond to these sharpening & detailing processes - all good learning to help with my own PP in future.
What am I Considering?
1) The Resolution & Print Size Issue
The question has been raised (on the UK forum) about possible print sizes - 'do you think you could print up to ......' ? My answer is not going to be straight. This is because I don't see this specific question as at all significant, or not exactly relevant to this particular comparison.
This question arises when thinking about a camera's 'resolution' - pixel dimensions and ppi figure. We make an assessment of the maximum print possible at a pre-determined ideal resolution figure (often 240ppi or 300ppi for print).
Both these cameras shoot photos at exactly the same technical resolution in these terms. The pixel figures and file sizes are identical. It struck me that discussions on quality as it related to perceptions of 'resolution' have not been reaching firm conclusions. That is why I started this - a new way to compare and assess the potential output, starting from the one true technical difference between the images - the blur or lack of blur at capture.
2) Image Content, Photo 'Genre' & Personal Workflow
The workflow tests and considerations of personal style and editing preferences then places this starting capture point in an individual context. In these conclusions I consider the likely effects of these personal choices based on what I've seen happen to the shots during the testing. What is really significant is the actual content of the image - the subject type, nature of the textures and density of the finer details. It is the question of how much enhancement this content can take to allow largescale output before any degradation resulting from this processing will become obvious at larger scales.
Also, photo genre will influence how you interpret the results of the tests here - the test image was basically a highly detailed still-life. The significance of what was seen under the testing and the processes applied will be different if you are thinking about your landscapes, portraits, sports shots, news reportage or macro work. For example, in the test image I saw the start of surface breakdown on smooth, plain textures. For me, the equivalent textures in my landscapes would be skys and clouds, so I need to think about how I would keep control of that, and whether it would detract from the overall perception of quality as it did in the end on the test image.
This is my 'New Take' - a new way to think about the significance of the AA filter and how you can assess the impact of that on your work along the line to final output, whether thats a huge print or a high quality compressed JPEG for the web.
3) The Cost Difference & Moire
Its hard not to factor these matters in to any equation, and again this will be a personal judgement. My own assessments will make the following assumptions -
For now the cost difference is significant. If the testing has indicated no likely technical benefit under any specific conditions, then I will factor in the cost issue as 'not worth the extra'. You may feel differently.
Balanced against any positive results is the risk of moire defects. I was genuinely surprised to see the effects of this on a fairly common type of ribbed surface in the test shot. Also I've seen it on a bird in a shot posted on the (UK) forum recently. I don't currently have any software to remove this problem, and it will be another concern during processing.
So for me the moire issue is a risk, and I will not discount it when making my choice. You may feel differently.
Finally, from now on I will stop referring to a 'K5II', of either version. This is because I will be using these conclusions in my future purchaing choice, which hopefully will be a new generation top-of-range Pentax DSLR, also offering the 'aa or no aa' options What My Tests Indicate to Me
1) The JPEG Preview
These embedded images were examined in PDCU's preview windows. They had been processed in the camera at capture point with the camera-set adjustments applied, which were identical apart from a single unit WB setting.
Viewed at 100% I was unable to distinguish anything between them. It is apparent that there is additional in-camera sharpening applied through the default camera algorithms on the K5II images. This has the effect of equalizing the shots in this respect. I don't shoot camera JPEGs, so there's no personal significance to this....
1
) Conclusion - based on this testing approach, if you shoot JPEGs and do not carry out extra PP work including sharpening or other detailing enhancement, you will not see any benefit to the 's' camera
- Choose AA filter model 2) Basic RAW conversion and Capture Sharpening adjustment in PDCU
With this process, all software settings were neutralised. However, by default, the K5II image very nearly equalised the 's' shot. This is apparently due to the camera profiles being interperted by PDCU and additional default sharpening being applied when the file was opened. A +1 user adjustment to the K5II shot made them almost identical.
2) Conclusion - for me this is significant. I do my RAW conversion in PDCU, I save out with a good bit of capture sharpening (+2 or +3 for most shots). I then carry on the workflow in other software, including significant detail enhancements for highly intricate scenes. For shots that do not require this extra work, (eg shots of the family) then the Capture Sharpening is as much as I often need.
So, if you do not need to (or don't like to) apply significant amounts of detail, fine micro-contrast and other sharpening-type enhancements to your images, you will not see significant benefit with the 's' model
- choose AA Filter model 3) Additional Definition & Detailing Enhancements - to larger output sizes
If following RAW conversion, including Capture Sharpening, you need to carry out further sharpening, definition or other type of clarity enhancements in PP, and these enhancements are applied to a wide range of textural surfaces and differing ranges of pixel radius, then there comes a point when the K5II image will let you down before the 's' version, particularly on smoother textures and larger areas of single tone. Also deep shadow areas will lose their depth more quickly.
3) Conclusion - If your shots demand these more intense processes, and large prints risk revealing these defects
- choose non AA filter ('s') model 4) Basic Photoshop-Style Sharpening
Finally I subjected the shots to progressivly higher amounts of fine edge sharpening (radius 0.7). Up to a very high level (K5IIs 200% - K5II 300%) each shot maintained quality acceptably. After this threshold the K5II image started to break down in smoother textural areas, flatter surfaces becoming grainy and rough.
Now, if your Photoshop-Style sharpening is limited to this approach (ie at a single radius setting), you are going to be hitting this limit with the K5II for your very detailed and intricate shots, unless you don't like them sharp. The only other choice would be to use a layers and masking-type approach to apply differing degrees of sharpening to different radius settings, thereby deliberately reducing the sharpening given to those more delicate softer areas. But this is a lot more work.
4) Conclusion - if your PS-style sharpening is applied globally to a single radius setting, and you need to apply high amounts to very detailed, inticate subjects
- choose non AA Filter 's' model
If you like to (or choose to) work more intricately and isolate different parts of your shots, reducing the intensity of sharpening on softer textures by use of layers and masking etc,
- choose AA filter model Finally - What Will I Choose?
Clearly, I have to make a balanced judgement call. My requirements fall within options 2) 3) & 4), depening on the type of image I've got.
My most critical shots are my landscapes, usually a mixture of dense and detailed ground and foliage textures, with smooth skys and clouds. Here I apply selections or masking when boosting sharpness and detail, so I'm not really worried about messing up my skys. The same applies to people shots, where I will isolate the skin textures, only applying fine detail enhancements to other parts of the shot. So the issue of smooth textures degrading under more intense sharpening techniques is not a worry for me.
Moire is a concern however. I do not currently have software to remove it, so it would need extra thought and effort. I would prefer not to have it as an extra concern.
Cost is hard to predict, as who know's how much the difference between the options will be in the future. But I'm not going to want to pay much more if I can't identify a clear and consistent advantage to me personally.
So the choice for me is clear. Weighting up all this, as a result of this testing, for me the best choice will be the camera that continues to include the AA filter.
Thats it! I love doing this sort of thing, and I hope the thinking and considerations behind it are informative and add to the range of opinions and experiences we can all draw on in making this new kind of technology choice.
Please keep the comments and feedback coming. I would be very glad to hear any thought about the tests or the conclusions! Or better still, alternative ways to test or think about it.
Nigel McGregor, June 2013, London UK
Originally posted by trumanusa 
Similar Threads 
