Photokit Sharpener II and non AA filter sensors?

[Jeff Kott]

I've been a PK Sharpener (and now PK Sharpener II)  user for many years and love the product.

Until now, I've never had a digital camera that did not have an AA filter, but I am about to receive a Sony A7/R (36 mp sensor with no AA filter).

Does the lack of an AA filter make capture sharpening unnecessary with A7/R, D800e and RX1r files?

If capture sharpening with files from these cameras is necessary, please explain why and which PK Sharpener II capture sharpening setting are recommended for these cameras.

Thanks in advance for you advice.

Jeff Kott
San Francisco

[bjanes]

I've been a PK Sharpener (and now PK Sharpener II)  user for many years and love the product.

Until now, I've never had a digital camera that did not have an AA filter, but I am about to receive a Sony A7/R (36 mp sensor with no AA filter).

Does the lack of an AA filter make capture sharpening unnecessary with A7/R, D800e and RX1r files?

If capture sharpening with files from these cameras is necessary, please explain why and which PK Sharpener II capture sharpening setting are recommended for these cameras.

Thanks in advance for you advice.

Jeff Kott
San Francisco

Jeff,

Theoretically, less capture sharpening would be needed for a camera like the 800e that lacks a low pass filter. In the original Real World Sharpening book (2007), Bruce Fraser discusses the strategies for cameras with and without low pass filters (pages 165-168). One would use the same radius for both instances, but the amount would have to be greater for the camera with the low pass filter.

I have the 800e and can confirm that sharpening is definitely necessary with this camera. I sometimes use Photokit II and adjust the sharpening amount at 100% viewing size. If the image appears oversharpened, I decrease the opacity somewhat. However, I now do most of my capture sharpening in ACR or LR. I don't know if the default is adjusted for the lack of a blur filter, but I usually increase the amount over the default with good results.

It would be interesting to learn what Jeff Schewe recommends.

Regards,

Bill

[Jeff Kott]

Bill,

Thanks for your quick reply. Using PK Sharpener has made me lazy/ignorant regarding sharpening, because I always get such great results using the plug in.

I hate to ask such a basic question, but would you please tell me what visual cues you are looking for on screen at 100% to tell you when you have the proper amount of capture sharpening?

Thanks!

Jeff

[Paul2660]

The nature of the Bayer pattern sensor even without an AA filter will still require sharpening.  You are still interpolating 2 colors for every color recorded by the sensor.  This alone will create a bit of blur.  In theory only the Foveon sensor would eliminate this since each color is captured by a separate sensor layer.

As I understand what Nikon did, the 800e still has the Low pass filter but has a some other filter installed to counteract the effects of it.   I shot both, 800 and 800e and still can't see much difference but kept the e.

With a Phase One back, no AA is bit more noticeable IMO.  You will see some aliasing on finer objects at times but Capture One can fix most of this.  Mainly the "Christmas Tree light" look from bayer pattern aliasing. 

I am a big user of Photokit, not as much Capture sharpening, as I will rely on either LR or Capture one for that, but their Creative Sharpening can really help.  Since I started using Focus Magic again, I am using Photokit less, just love the look
from the files processed out by Focus Magic. 

Enjoy the Sony, should be a great camera.

Paul Caldwell

[Schewe]

It would be interesting to learn what Jeff Schewe recommends.

Well, I no longer use PKS2 to do capture sharpening as I do it in Lightroom. I also don't do output sharpening other than in Lightroom. I do go into Photoshop a lot for PKS2 creative sharpening...

The best suggestion I can offer is to do some tests with your new camera and see what you think. You have the ability to tweak the final opacity...the aim point for capture sharpening is to achieve a good image sharpness at 100% zoom without sharpening defects.

[bjanes]

Bill,

Thanks for your quick reply. Using PK Sharpener has made me lazy/ignorant regarding sharpening, because I always get such great results using the plug in.

I hate to ask such a basic question, but would you please tell me what visual cues you are looking for on screen at 100% to tell you when you have the proper amount of capture sharpening?

I see that Jeff Schewe has already replied. One basically adjusts by eye at 100% magnification until the best looking image is obtained. It is important to avoid oversharpening artifacts. With PKS2 one can select the sharpening edge effect: auto edge, superfine edge, medium edge, etc. This is basically adjusting the sharpening radius, and preview allows one to choose the most appropriate effect for a given image. The sharpening amount is adjusted after running the plugin by adjusting the opacity of the sharpening layer.

With ACR and LR one must set the parameters manually, although LR does have some presets. Jeff Schewe outlines the basics in his recent book, The Digital Negative. Since the capture and output sharpening of PKS2 is incorporated into Lightroom, many users employ PKS2 mainly for creative sharpening.

Bill

[Bart_van_der_Wolf]

I've been a PK Sharpener (and now PK Sharpener II)  user for many years and love the product.

Until now, I've never had a digital camera that did not have an AA filter, but I am about to receive a Sony A7/R (36 mp sensor with no AA filter).

Does the lack of an AA filter make capture sharpening unnecessary with A7/R, D800e and RX1r files?

Hi Jeff,

Capture sharpening is always necessary, if you want to restore the original subject sharpness instead of the blur that the Capture process created.

If capture sharpening with files from these cameras is necessary, please explain why and which PK Sharpener II capture sharpening setting are recommended for these cameras.

The Capture process is inherently blurry, but the amount and degree varies, mostly with aperture value (=diffraction). Residual lens aberrations and defocus are other causes, as is the area sampling process and demosaicing.

As others have already indicated, many use the sharpening controls of e.g. Lightroom or ACR. The problem with those is that there is only a limited amount of feedback which indicates whether you've applied the correct(!) radius on which all other parameters should be based.

To satisfy one's curiosity, or to accurately characterize one's lens+camera combinations, you could use this tool.

Cheers,
Bart

[Bart_van_der_Wolf]

I hate to ask such a basic question, but would you please tell me what visual cues you are looking for on screen at 100% to tell you when you have the proper amount of capture sharpening?

Hi Jeff,

It depends on the subject matter in the image whether you'll be able to use visual clues to judge the optimal setting of radius and amount (in that order). If there is a sharp (in the focus plane) edge at a slant, it will be possible to check for the formation of halo artifacts. Without such clues, you can use settings determined earlier, because they will mostly vary only by aperture used. Optimal Capture sharpening is very much a hardware correction, not subject related.

Cheers,
Bart

[Jeff Kott]

If I was a Lightroom user, I'd follow the suggestion to sharpen in Lightroom but I use Capture One as my raw converter. I've found I prefer the results that I get converting my file with C1 sharpening turned off and then opening the converted Tiff in Photoshop and using PKS2.

Many thanks for the great responses and ideas on different alternatives to try with my sharpening work flow!

Paul, I see you're also a C1 user but prefer Focus Magic to PKS2 and the C1 sharpening, so I'll give it a try and see if I can see what you're seeing.

Bart, your tool looks like a great answer to my question about how you can tell when you've got the appropriate amount of sharpening, so I'm going to give that a try also.

[bjanes]

Hi Jeff,

It depends on the subject matter in the image whether you'll be able to use visual clues to judge the optimal setting of radius and amount (in that order). If there is a sharp (in the focus plane) edge at a slant, it will be possible to check for the formation of halo artifacts. Without such clues, you can use settings determined earlier, because they will mostly vary only by aperture used. Optimal Capture sharpening is very much a hardware correction, not subject related.

At this point for interested readers it should be noted that there is a divergence of opinion regarding the choice of a proper radius for sharpening between the Bart and the Schewe-Fraser approaches for sharpening. Both agree that the radius is the most critical factor. In Bart's approach, one chooses the radius according to a scientific measurement of the blur radius, where a larger radius is necessary at smaller apertures (larger f/numbers) to take into account the effects of diffraction. In this approach, subject matter is not taken into account. In the Schewe-Fraser approach, the radius is chosen in part according to image content. One uses a smaller radius for images with high frequency image detail such as landscapes and a larger radius for images with lower frequency detail such as portraits, where it is not desirable to image each pore on the subjects nose. The assumption here is that one can selectively enhance certain image frequencies. Further discussion on this matter would be welcome.

Regards,

Bill

[Fine_Art]

At this point for interested readers it should be noted that there is a divergence of opinion regarding the choice of a proper radius for sharpening between the Bart and the Schewe-Fraser approaches for sharpening. Both agree that the radius is the most critical factor. In Bart's approach, one chooses the radius according to a scientific measurement of the blur radius, where a larger radius is necessary at smaller apertures (larger f/numbers) to take into account the effects of diffraction. In this approach, subject matter is not taken into account. In the Schewe-Fraser approach, the radius is chosen in part according to image content. One uses a smaller radius for images with high frequency image detail such as landscapes and a larger radius for images with lower frequency detail such as portraits, where it is not desirable to image each pore on the subjects nose. The assumption here is that one can selectively enhance certain image frequencies. Further discussion on this matter would be welcome.

Regards,

Bill

Bill,

I do not thing these are competing methods. The Schewe-Fraser method is asking what do you need from the image. The Bart method is showing how to get optimal sharpening if it is needed, without making a judgement about the if. That is up to the photographer.

[Paul2660]

Jeff:

You bring up a good point on C1, which I use on phase one files, D800e, and Fuji files.  C1 loads a default capture sharpening which varies by camera type.  I tend to use that amount.  Capture One also loads a set amount of noise reduction which I tend to reduce.  LR on the other hand loads the same sharpening defaults for everything and I will vary it by camera type. 

With c1 conversions raw conversions I tend add some extra sharpening with focus magic or photokit creative sharpening again depending on the image.

Paul Caldwell

[Bart_van_der_Wolf]

I do not think these are competing methods.

Hi Arthur,

Perhaps not competing, but totally different in approach to image quality.

The Schewe-Fraser method is asking what do you need from the image.

Yes, but IMHO, that's what Creative and Output sharpening is for. Creative sharpening is also much better done with more appropriate tools (spatial frequency targeted and local contrast adjusted), later in the chain of processing activities.

The Bart method is showing how to get optimal sharpening if it is needed, without making a judgement about the if.

That's right. Just like Raw conversion, Capture sharpening is all about getting an optimal (artifact free, and free of issues that will negatively impact good postprocessing capability) RGB baseline from our Raw capture. That Capture is by definition blurred due to the Capture process itself. The deconvolution sharpening will restore actual scene sharpness, which also improves color accuracy at the pixel level. Using the wrong radius will not do that!

The lens specific optimal radius can vary quite a bit with aperture used, as illustrated by this analysis of the actual amount of blur in optimally (focus rail) focused images:

The required radius is very much hardware (Sensel pitch, Lens+Aperture , plus Raw-conversion) driven as can be seen (EXIF data is available to improve the Raw converter defaults).

That's also why the sharpening dialog needs to be improved, and give better feedback about the optimal radius, or as a minimum use more sensible defaults. Human vision is just not good enough to do that objectively by eye for many types of subject matter.

It's just like noise reduction or CA correction, it too should be addressed early in the chain of events, preferably in linear gamma space before demosaicing and gamma adjustments break the independent random structure that the noise reduction is based on.

Cheers,
Bart

[hjulenissen]

The nature of the Bayer pattern sensor even without an AA filter will still require sharpening.  You are still interpolating 2 colors for every color recorded by the sensor.  This alone will create a bit of blur.  In theory only the Foveon sensor would eliminate this since each color is captured by a separate sensor layer.

The Foveon sensor (even without AA filter) will still have an effective sensel area. This area represents a spatial filter with a particular lowpass response (sin(x)/x for a 1-d sensor) that does some attenuation of high frequencies below Nyquist.

If you had a achromatic camera with a perfect pulse lens PSF, exactly no camera/scene movement, a sensel fill-rate approaching zero (effectively a point-sampler), and an OLPF approaching a lanczos 2/3 shape (including "negative light" contribution), then I would be inclined to agree that capture blurring was eliminated and no capture sharpening needed.

-h

[bjanes]

Perhaps not competing, but totally different in approach to image quality.

Yes, but IMHO, that's what Creative and Output sharpening is for. Creative sharpening is also much better done with more appropriate tools (spatial frequency targeted and local contrast adjusted), later in the chain of processing activities.

That's right. Just like Raw conversion, Capture sharpening is all about getting an optimal (artifact free, and free of issues that will negatively impact good postprocessing capability) RGB baseline from our Raw capture. That Capture is by definition blurred due to the Capture process itself. The deconvolution sharpening will restore actual scene sharpness, which also improves color accuracy at the pixel level. Using the wrong radius will not do that!

Although the term "Capture Sharpening" has become ingrained in photographic parlance, with the application of deconvolution algorithms it might be better to use the term "image restoration" to denote this process. Application of the unsharp mask does not restore data lost in the capture process, but merely creates the impression of sharpness by the introduction of sharpening halos that accentuate edge contrast. As Roger Clark explains in the above linked article, deconvolution algorithms actually restore details lost in the capture process.  This results in an image that more closely approximates the original scene and is relatively free of artifacts. The resulting master image can be further processed by creative and output sharpening to emphasize certain image frequencies.

Bill

[hjulenissen]

Although the term "Capture Sharpening" has become ingrained in photographic parlance, with the application of deconvolution algorithms it might be better to use the term "image restoration" to denote this process. Application of the unsharp mask does not restore data lost in the capture process, but merely creates the impression of sharpness by the introduction of sharpening halos that accentuate edge contrast. As Roger Clark explains in the above linked article, deconvolution algorithms actually restore details lost in the capture process.  This results in an image that more closely approximates the original scene and is relatively free of artifacts. The resulting master image can be further processed by creative and output sharpening to emphasize certain image frequencies.

Bill

This might seem pedantic, but in my view, no algorithm can "restore lost data". The recorded data can be re-arranged in such a way as to more closely match the (most likely) true scene. If the camera system have a systematic (predictable) loss of high spatial-frequency data, then one can attempt an approximate/regularized inversion of this degradation to come "closer" to the ideal data. This can be done in some mathematical sense (e.g. Least Squared) or subjective sense.

I would think that USM does re-shape the image data in a form closer to the ideal image for cases where the implicit degradation model of USM match that of the camera closely. Less so in cases where it does not match. I believe that what deconvolution fundamentally brings to the table is a variable degradation model (PSF) that is either estimated offline or blindly.

-h

[Bart_van_der_Wolf]

This might seem pedantic, but in my view, no algorithm can "restore lost data".

Hi,

Obviously one cannot really restore lost data (although an educated guess can go a long way, as e.g. demosaicing proves), but one can restore visually lost detail. So the seemingly lost data can be restored to a large extent (limited by recorded precision, quantization error, PSF accuracy).

All one has to do is look at the examples here to understand that not all is lost that seems lost. USM of the blurred image is no way as successful in restoring real resolution.

Cheers,
Bart

[Fine_Art]

It's just like noise reduction or CA correction, it too should be addressed early in the chain of events, preferably in linear gamma space before demosaicing and gamma adjustments break the independent random structure that the noise reduction is based on.

This gets to the heart of the matter. While a bit off topic, we are looking for a method to get the best results. Maybe that is provided by a black box software solution, maybe we can go through the process ourselves with a set of basic tools.

So what is the best sequence? Remove noise first? Then deconvolve, then demosaic, then gamma, then the ususal assortment of raw tools?

[digitaldog]

Page 8 of the PKS manual:

For best results, we recommend starting out with unsharpened files (turn off any sharpening in the scanner or capture software), and using PhotoKit SHARPENER to apply all sharpening.

So the question now becomes, do you want to apply capture sharpening at the raw acquisition stage or use PKS II? Depends on your raw processor. Jeff has answered about ACR/LR's sharpening and workflow. You'll have to test this for yourself. I'd vastly prefer to do this at the raw/parametric stage like Jeff and do so now as well. The workflow, the data, the flexibility is all over the map and much depends on your ability to sharpen as early as possible. I also get rendered images from time to time I have to work on and often, they need a dose of PKS Capture Sharpening.

[Bart_van_der_Wolf]

Page 8 of the PKS manual:So the question now becomes, do you want to apply capture sharpening at the raw acquisition stage or use PKS II?

Hi Andrew,

After demosaicing, the data is no longer Raw, but it is blurry. So both PKS and deconvolution are faced with the same challenge, and offer to either increase local contrast (=PKS) or restore resolution (=deconvolution).

Cheers,
Bart