Current X3F-processing best practice?

[Scottelly]

Here's my result from X3F Wrapper, DNG directly processed in Camera Raw. Manually white balanced on the snow and using a profile home-made with DNG Profile Editor.


DNG from X3F Wrapper 0.56 Beta 3, developed in ACR

Georg, I like your version more, but it does seem to be sharpened more, which causes jaggies on the edge of the flag on that lamp post just right of middle.

But they are not true "jaggies" - 800% pixel-peeping reveals that the edges in that flag are properly anti-aliased, just like Lumo's "perfect pixel" (1.2.2) here:

http://www.falklumo.com/lumolabs/articles/sharpness/index.html

See his "hard pixel" (1.2.1) for true jaggies.

Interesting Ted. Thanks.

Here is what I'm talking about: <examples>

Thanks for the examples but I had already down-loaded Georg's image and viewed it at 800%.

As you can probably tell, you need to not only view at full-size, but then you need to zoom in, so you can see the "jaggies" that I saw, which maybe some people would call stair-stepping. I think the sharpening enhances them basically every time, which is one reason I don't like sharpening.

A matter of terminology, I guess, Scott.

Lumo's hard jaggies or stair-stepping:



Lumo's aliased jaggies or stair-stepping:



The point here is that a perfectly anti-aliased edge is correct and is the best you can get with a digital image. The edges in Georg's flag rendition look much more like the second than they do the first. To smooth such an edge to "get rid of the jaggies" would do it a disservice.

Well Ted, certainly Georg's version of the image looks "sharper" (probably because of more sharpening done in post-processing). I'm not so sure that NOT sharpening so much would really be doing the image a disservice, but I do like the look of his image better, so maybe you are right.

--
Scott Barton Kennelly

Big Print Photos - Photography by Scott Barton Kennelly

Landscape and nature photography, photographic prints by Scott Barton Kennelly, an Australian photographer living in south Florida.

www.bigprintphotos.com

 

[xpatUSA]

Here's my result from X3F Wrapper, DNG directly processed in Camera Raw. Manually white balanced on the snow and using a profile home-made with DNG Profile Editor.


DNG from X3F Wrapper 0.56 Beta 3, developed in ACR

Georg, I like your version more, but it does seem to be sharpened more, which causes jaggies on the edge of the flag on that lamp post just right of middle.

But they are not true "jaggies" - 800% pixel-peeping reveals that the edges in that flag are properly anti-aliased, just like Lumo's "perfect pixel" (1.2.2) here:

http://www.falklumo.com/lumolabs/articles/sharpness/index.html

See his "hard pixel" (1.2.1) for true jaggies.

Interesting Ted. Thanks.

Here is what I'm talking about: <examples>

Thanks for the examples but I had already down-loaded Georg's image and viewed it at 800%.

As you can probably tell, you need to not only view at full-size, but then you need to zoom in, so you can see the "jaggies" that I saw, which maybe some people would call stair-stepping. I think the sharpening enhances them basically every time, which is one reason I don't like sharpening.

A matter of terminology, I guess, Scott.

Lumo's hard jaggies or stair-stepping:



Lumo's aliased jaggies or stair-stepping:



The point here is that a perfectly anti-aliased edge is correct and is the best you can get with a digital image. The edges in Georg's flag rendition look much more like the second than they do the first. To smooth such an edge to "get rid of the jaggies" would do it a disservice.

Well Ted, certainly Georg's version of the image looks "sharper" (probably because of more sharpening done in post-processing).

Whenever Scott starts a response with "Well," there's often a rebuttal to follow. Scott continues to say "sharper" and "more sharpening" in spite of my claim of it being about right.

Time to bring in the big guns - QuickMTF and Falk Lumo, whose paper is being studiously ignored :-|

I found an edge close enough to the flag to be significant. Here is living proof that Georg's image is SLIGHTLY SOFT:



Lumo tells us that a perfect sharp edge response is 1.26px rise from 10-90%; Georg's edge above has a 2.70px rise which makes it "OK" but a bit SOFT.

I'm not so sure that NOT sharpening so much would really be doing the image a disservice ...

(sigh) - a nice vague double negative ... I actually said "To smooth such an edge to get rid of the jaggies would do it a disservice."

but I do like the look of his image better, so maybe you are right.

No "maybe" about it, IMNSHO. :-D

--
... sound of biplane in flames spirally down somewhere in Floorduh ... ;-)

 

[geni1105]

I found an edge close enough to the flag to be significant. Here is living proof that Georg's image is SLIGHTLY SOFT:

Lumo tells us that a perfect sharp edge response is 1.26px rise from 10-90%; Georg's edge above has a 2.70px rise which makes it "OK" but a bit SOFT.

Just re-opened the image in ACR, and I think you are correct: I used my ACR defaults for DNGs, which are at 30 / 1.0 / 25 (sharpen / radius / detail), whereas Adobe's are 40 / 1.0 / 25.

I guess Adobe's defaults looked overdone to me, so I tuned it down a bit some time ago.

 

[geni1105]

Regarding green edges and field flattening: as mentioned by EEvan above, you need to first activate "Correct Merrill color cast" in the Wrapper's configuration dialog.

…

(If you have activated color cast correction and still see the green edges in the Kalpanika created DNG, your raw converter is not fully implementing the DNG specification, in particular not evaluating the Opcode3 tags.

Regards, Georg

Hi Georg,

What is the source of the opcode files? I was curious to see them "in person", so I extracted them from x3f_wrapper's Opcodelist3 files, and, um, they are not perfect.

Here are the color cast maps from those opcodes with contrast heavily emphasized. I used a formula like this: ((original_value - 1)*some_big_number+0.5).

DP1M images by aperture, as a .gif animation. That center circle is not centered and jumps around.

All other cameras opcode files are exactly the same per aperture, so no animation.

Poor DP2M must have some retina problems.

DP3M's map only correct vignetting, apparently colors are not corrected. Also, interesting lines are visible.

Maybe we should make new color cast correction files. If only I knew how to make a perfectly homogenous light source…

I made new ones for the DP2M and DP3M, see https://www.dropbox.com/sh/vd601dndp465oq9/AACR7B1D91oiDu3ihvTkTbgKa?dl=0 , would you be so kind and evaluate those too?

I simply went outside on an overcast day, put a white cardboard on the ground and took the photos from about 50 cm above, taking care not to cast a shadow on the cardboard of course. This may not be perfectly homogeneous, but I think it's ok.

I need to confess, it seems that I made a stupid mistake with the old sets, misinterpreting the functionality of LIghtroom's flat-fielding plugin: I applied it to the full set of 32 images (16 test images, 16 reference images) at once, assuming that the plugin would cleverly pair the images shot at the same aperture. Turns out it does not, instead apparently it averages the reference shots and applies the result to all test images - hence the same opcode for all apertures. My bad, embarassing ...

This time I took care to take just the two images at a time, and repeated 16x.
Still surprising that the averaged opcodes do a decent job, at least it was not immediately visible to me that something is wrong ...

 

[Scottelly]

Here's my result from X3F Wrapper, DNG directly processed in Camera Raw. Manually white balanced on the snow and using a profile home-made with DNG Profile Editor.


DNG from X3F Wrapper 0.56 Beta 3, developed in ACR

Georg, I like your version more, but it does seem to be sharpened more, which causes jaggies on the edge of the flag on that lamp post just right of middle.

But they are not true "jaggies" - 800% pixel-peeping reveals that the edges in that flag are properly anti-aliased, just like Lumo's "perfect pixel" (1.2.2) here:

http://www.falklumo.com/lumolabs/articles/sharpness/index.html

See his "hard pixel" (1.2.1) for true jaggies.

Interesting Ted. Thanks.

Here is what I'm talking about: <examples>

Thanks for the examples but I had already down-loaded Georg's image and viewed it at 800%.

As you can probably tell, you need to not only view at full-size, but then you need to zoom in, so you can see the "jaggies" that I saw, which maybe some people would call stair-stepping. I think the sharpening enhances them basically every time, which is one reason I don't like sharpening.

A matter of terminology, I guess, Scott.

Lumo's hard jaggies or stair-stepping:



Lumo's aliased jaggies or stair-stepping:



The point here is that a perfectly anti-aliased edge is correct and is the best you can get with a digital image. The edges in Georg's flag rendition look much more like the second than they do the first. To smooth such an edge to "get rid of the jaggies" would do it a disservice.

Well Ted, certainly Georg's version of the image looks "sharper" (probably because of more sharpening done in post-processing).

Whenever Scott starts a response with "Well," there's often a rebuttal to follow. Scott continues to say "sharper" and "more sharpening" in spite of my claim of it being about right.

Time to bring in the big guns - QuickMTF and Falk Lumo, whose paper is being studiously ignored :-|

I found an edge close enough to the flag to be significant. Here is living proof that Georg's image is SLIGHTLY SOFT:



Lumo tells us that a perfect sharp edge response is 1.26px rise from 10-90%; Georg's edge above has a 2.70px rise which makes it "OK" but a bit SOFT.

I'm not so sure that NOT sharpening so much would really be doing the image a disservice ...

(sigh) - a nice vague double negative ... I actually said "To smooth such an edge to get rid of the jaggies would do it a disservice."

but I do like the look of his image better, so maybe you are right.

No "maybe" about it, IMNSHO. :-D

I think maybe the edge you picked is not the best example of a sharp edge Ted, since the bark of that tree might be a bit too bumpy to make a sharp edge. I'm not able to view the image on a big screen right now though, so I guess I could be wrong.

--
Scott Barton Kennelly
https://www.bigprintphotos.com/

 

[Scottelly]

Regarding green edges and field flattening: as mentioned by EEvan above, you need to first activate "Correct Merrill color cast" in the Wrapper's configuration dialog.

…

(If you have activated color cast correction and still see the green edges in the Kalpanika created DNG, your raw converter is not fully implementing the DNG specification, in particular not evaluating the Opcode3 tags.

Regards, Georg

Hi Georg,

What is the source of the opcode files? I was curious to see them "in person", so I extracted them from x3f_wrapper's Opcodelist3 files, and, um, they are not perfect.

Here are the color cast maps from those opcodes with contrast heavily emphasized. I used a formula like this: ((original_value - 1)*some_big_number+0.5).

DP1M images by aperture, as a .gif animation. That center circle is not centered and jumps around.

All other cameras opcode files are exactly the same per aperture, so no animation.

Poor DP2M must have some retina problems.

DP3M's map only correct vignetting, apparently colors are not corrected. Also, interesting lines are visible.

Maybe we should make new color cast correction files. If only I knew how to make a perfectly homogenous light source…

I made new ones for the DP2M and DP3M, see https://www.dropbox.com/sh/vd601dndp465oq9/AACR7B1D91oiDu3ihvTkTbgKa?dl=0 , would you be so kind and evaluate those too?

I simply went outside on an overcast day, put a white cardboard on the ground and took the photos from about 50 cm above, taking care not to cast a shadow on the cardboard of course. This may not be perfectly homogeneous, but I think it's ok.

I need to confess, it seems that I made a stupid mistake with the old sets, misinterpreting the functionality of LIghtroom's flat-fielding plugin: I applied it to the full set of 32 images (16 test images, 16 reference images) at once, assuming that the plugin would cleverly pair the images shot at the same aperture. Turns out it does not, instead apparently it averages the reference shots and applies the result to all test images - hence the same opcode for all apertures. My bad, embarassing ...

This time I took care to take just the two images at a time, and repeated 16x.
Still surprising that the averaged opcodes do a decent job, at least it was not immediately visible to me that something is wrong ...

Is this something that you will use to update Kalpanica with? It seems Kalpanica is getting updates from time to time. I think that's very cool.

I need to figure out the best/easiest/quickest ways to remove color casts caused by my old lenses. Do you have any suggestions where I could find such information?

--
Scott Barton Kennelly

Big Print Photos - Photography by Scott Barton Kennelly

Landscape and nature photography, photographic prints by Scott Barton Kennelly, an Australian photographer living in south Florida.

www.bigprintphotos.com

 

[Scottelly]

I found an edge close enough to the flag to be significant. Here is living proof that Georg's image is SLIGHTLY SOFT:

Lumo tells us that a perfect sharp edge response is 1.26px rise from 10-90%; Georg's edge above has a 2.70px rise which makes it "OK" but a bit SOFT.

Just re-opened the image in ACR, and I think you are correct: I used my ACR defaults for DNGs, which are at 30 / 1.0 / 25 (sharpen / radius / detail), whereas Adobe's are 40 / 1.0 / 25.

I guess Adobe's defaults looked overdone to me, so I tuned it down a bit some time ago.

Thanks for the info! So there is indeed some significant sharpening done to your version. I don't think it was too much though, since I had to view at 100% and then magnify that view in order to see the stair-stepping (and I didn't notice any obvious sharpening halos), but I definitely think you didn't "under-sharpen" the image. It looks plenty "sharp" to me, despite what Ted says about it being slightly soft.

--
Scott Barton Kennelly
https://www.bigprintphotos.com/

 

[xpatUSA]

Here's my result from X3F Wrapper, DNG directly processed in Camera Raw. Manually white balanced on the snow and using a profile home-made with DNG Profile Editor.


DNG from X3F Wrapper 0.56 Beta 3, developed in ACR

Georg, I like your version more, but it does seem to be sharpened more, which causes jaggies on the edge of the flag on that lamp post just right of middle.

But they are not true "jaggies" - 800% pixel-peeping reveals that the edges in that flag are properly anti-aliased, just like Lumo's "perfect pixel" (1.2.2) here:

http://www.falklumo.com/lumolabs/articles/sharpness/index.html

See his "hard pixel" (1.2.1) for true jaggies.

Interesting Ted. Thanks.

Here is what I'm talking about: <examples>

Thanks for the examples but I had already down-loaded Georg's image and viewed it at 800%.

As you can probably tell, you need to not only view at full-size, but then you need to zoom in, so you can see the "jaggies" that I saw, which maybe some people would call stair-stepping. I think the sharpening enhances them basically every time, which is one reason I don't like sharpening.

A matter of terminology, I guess, Scott.

Lumo's hard jaggies or stair-stepping:



Lumo's aliased jaggies or stair-stepping:



The point here is that a perfectly anti-aliased edge is correct and is the best you can get with a digital image. The edges in Georg's flag rendition look much more like the second than they do the first. To smooth such an edge to "get rid of the jaggies" would do it a disservice.

Well Ted, certainly Georg's version of the image looks "sharper" (probably because of more sharpening done in post-processing).

Whenever Scott starts a response with "Well," there's often a rebuttal to follow. Scott continues to say "sharper" and "more sharpening" in spite of my claim of it being about right.

Time to bring in the big guns - QuickMTF and Falk Lumo, whose paper is being studiously ignored :-|

I found an edge close enough to the flag to be significant. Here is living proof that Georg's image is SLIGHTLY SOFT:



Lumo tells us that a perfect sharp edge response is 1.26px rise from 10-90%; Georg's edge above has a 2.70px rise which makes it "OK" but a bit SOFT.

I'm not so sure that NOT sharpening so much would really be doing the image a disservice ...

(sigh) - a nice vague double negative ... I actually said "To smooth such an edge to get rid of the jaggies would do it a disservice."

but I do like the look of his image better, so maybe you are right.

No "maybe" about it, IMNSHO. :-D

It would be interesting to know what settings were used (particularly the Sharpness setting).

Except that I am only talking about the quality of the image itself, not how it was obtained.

--
WYSINWIG: what you see is not what I got.

 

[xpatUSA]

Here's my result from X3F Wrapper, DNG directly processed in Camera Raw. Manually white balanced on the snow and using a profile home-made with DNG Profile Editor.


DNG from X3F Wrapper 0.56 Beta 3, developed in ACR

Georg, I like your version more, but it does seem to be sharpened more, which causes jaggies on the edge of the flag on that lamp post just right of middle.

But they are not true "jaggies" - 800% pixel-peeping reveals that the edges in that flag are properly anti-aliased, just like Lumo's "perfect pixel" (1.2.2) here:

http://www.falklumo.com/lumolabs/articles/sharpness/index.html

See his "hard pixel" (1.2.1) for true jaggies.

Interesting Ted. Thanks.

Here is what I'm talking about: <examples>

Thanks for the examples but I had already down-loaded Georg's image and viewed it at 800%.

As you can probably tell, you need to not only view at full-size, but then you need to zoom in, so you can see the "jaggies" that I saw, which maybe some people would call stair-stepping. I think the sharpening enhances them basically every time, which is one reason I don't like sharpening.

A matter of terminology, I guess, Scott.

Lumo's hard jaggies or stair-stepping:



Lumo's aliased jaggies or stair-stepping:



The point here is that a perfectly anti-aliased edge is correct and is the best you can get with a digital image. The edges in Georg's flag rendition look much more like the second than they do the first. To smooth such an edge to "get rid of the jaggies" would do it a disservice.

Well Ted, certainly Georg's version of the image looks "sharper" (probably because of more sharpening done in post-processing).

Whenever Scott starts a response with "Well," there's often a rebuttal to follow. Scott continues to say "sharper" and "more sharpening" in spite of my claim of it being about right.

Time to bring in the big guns - QuickMTF and Falk Lumo, whose paper is being studiously ignored :-|

I found an edge close enough to the flag to be significant. Here is living proof that Georg's image is SLIGHTLY SOFT:



Lumo tells us that a perfect sharp edge response is 1.26px rise from 10-90%; Georg's edge above has a 2.70px rise which makes it "OK" but a bit SOFT.

I'm not so sure that NOT sharpening so much would really be doing the image a disservice ...

(sigh) - a nice vague double negative ... I actually said "To smooth such an edge to get rid of the jaggies would do it a disservice."

but I do like the look of his image better, so maybe you are right.

No "maybe" about it, IMNSHO. :-D

I think maybe the edge you picked is not the best example of a sharp edge Ted, since the bark of that tree might be a bit too bumpy to make a sharp edge. I'm not able to view the image on a big screen right now though, so I guess I could be wrong.

Feel free to pick out a better example Scott. You should not need a big screen to do that, I found my edge with a mere 24" ...

--
WYSINWIG: what you see is not what I got.

 

[xpatUSA]

I found an edge close enough to the flag to be significant. Here is living proof that Georg's image is SLIGHTLY SOFT:

Lumo tells us that a perfect sharp edge response is 1.26px rise from 10-90%; Georg's edge above has a 2.70px rise which makes it "OK" but a bit SOFT.

Just re-opened the image in ACR, and I think you are correct: I used my ACR defaults for DNGs, which are at 30 / 1.0 / 25 (sharpen / radius / detail), whereas Adobe's are 40 / 1.0 / 25.

I guess Adobe's defaults looked overdone to me, so I tuned it down a bit some time ago.

Thanks for the info! So there is indeed some significant sharpening done to your version. I don't think it was too much though, since I had to view at 100% and then magnify that view in order to see the stair-stepping (and I didn't notice any obvious sharpening halos), but I definitely think you didn't "under-sharpen" the image. It looks plenty "sharp" to me, despite what Ted says about it being slightly soft.

Like the Energizer bunny, Scott, still going and going ...

How sharp it looks is not the same as how sharp it is.

"it all depends on what the meaning of 'is' is ..."

 

[geni1105]

Regarding green edges and field flattening: as mentioned by EEvan above, you need to first activate "Correct Merrill color cast" in the Wrapper's configuration dialog.

…

(If you have activated color cast correction and still see the green edges in the Kalpanika created DNG, your raw converter is not fully implementing the DNG specification, in particular not evaluating the Opcode3 tags.

Regards, Georg

Hi Georg,

What is the source of the opcode files? I was curious to see them "in person", so I extracted them from x3f_wrapper's Opcodelist3 files, and, um, they are not perfect.

Here are the color cast maps from those opcodes with contrast heavily emphasized. I used a formula like this: ((original_value - 1)*some_big_number+0.5).

DP1M images by aperture, as a .gif animation. That center circle is not centered and jumps around.

All other cameras opcode files are exactly the same per aperture, so no animation.

Poor DP2M must have some retina problems.

DP3M's map only correct vignetting, apparently colors are not corrected. Also, interesting lines are visible.

Maybe we should make new color cast correction files. If only I knew how to make a perfectly homogenous light source…

I made new ones for the DP2M and DP3M, see https://www.dropbox.com/sh/vd601dndp465oq9/AACR7B1D91oiDu3ihvTkTbgKa?dl=0 , would you be so kind and evaluate those too?

I simply went outside on an overcast day, put a white cardboard on the ground and took the photos from about 50 cm above, taking care not to cast a shadow on the cardboard of course. This may not be perfectly homogeneous, but I think it's ok.

I need to confess, it seems that I made a stupid mistake with the old sets, misinterpreting the functionality of LIghtroom's flat-fielding plugin: I applied it to the full set of 32 images (16 test images, 16 reference images) at once, assuming that the plugin would cleverly pair the images shot at the same aperture. Turns out it does not, instead apparently it averages the reference shots and applies the result to all test images - hence the same opcode for all apertures. My bad, embarassing ...

This time I took care to take just the two images at a time, and repeated 16x.
Still surprising that the averaged opcodes do a decent job, at least it was not immediately visible to me that something is wrong ...

Is this something that you will use to update Kalpanica with? It seems Kalpanica is getting updates from time to time. I think that's very cool.

I need to figure out the best/easiest/quickest ways to remove color casts caused by my old lenses. Do you have any suggestions where I could find such information?

How to update the official version of X3F Wrapper (the Kalpanika GUI, so to speak) on github is not obvious for me, since the latest version 0.57 does not work well with Merrills and branching from 0.56 might be confusing for users.

But you can take the opcodes from the dropbox folder, unzip and paste them into X3F Wrapper (navigate to the opcodes folder of the application - on Windows it's directly visible, on Mac right-click on the application and select "show package contents").

To remove color casts, you might also create opcodes for your lenses - I have described the procedure here https://www.dpreview.com/forums/post/58993342 . Just beware in step 4 *not* to select all images at once, but in pairs and repeat the flat-fielding for each pair. Instead of step 6, you can also paste your new opcodes into X3F Wrapper. The procedure only works for fixed focal lenses, unfortunately, not for zooms where the color cast probably depends on the focal length.

 

[Scottelly]

Here's my result from X3F Wrapper, DNG directly processed in Camera Raw. Manually white balanced on the snow and using a profile home-made with DNG Profile Editor.


DNG from X3F Wrapper 0.56 Beta 3, developed in ACR

Georg, I like your version more, but it does seem to be sharpened more, which causes jaggies on the edge of the flag on that lamp post just right of middle.

But they are not true "jaggies" - 800% pixel-peeping reveals that the edges in that flag are properly anti-aliased, just like Lumo's "perfect pixel" (1.2.2) here:

http://www.falklumo.com/lumolabs/articles/sharpness/index.html

See his "hard pixel" (1.2.1) for true jaggies.

Interesting Ted. Thanks.

Here is what I'm talking about: <examples>

Thanks for the examples but I had already down-loaded Georg's image and viewed it at 800%.

As you can probably tell, you need to not only view at full-size, but then you need to zoom in, so you can see the "jaggies" that I saw, which maybe some people would call stair-stepping. I think the sharpening enhances them basically every time, which is one reason I don't like sharpening.

A matter of terminology, I guess, Scott.

Lumo's hard jaggies or stair-stepping:



Lumo's aliased jaggies or stair-stepping:



The point here is that a perfectly anti-aliased edge is correct and is the best you can get with a digital image. The edges in Georg's flag rendition look much more like the second than they do the first. To smooth such an edge to "get rid of the jaggies" would do it a disservice.

Well Ted, certainly Georg's version of the image looks "sharper" (probably because of more sharpening done in post-processing).

Whenever Scott starts a response with "Well," there's often a rebuttal to follow. Scott continues to say "sharper" and "more sharpening" in spite of my claim of it being about right.

Time to bring in the big guns - QuickMTF and Falk Lumo, whose paper is being studiously ignored :-|

I found an edge close enough to the flag to be significant. Here is living proof that Georg's image is SLIGHTLY SOFT:



Lumo tells us that a perfect sharp edge response is 1.26px rise from 10-90%; Georg's edge above has a 2.70px rise which makes it "OK" but a bit SOFT.

I'm not so sure that NOT sharpening so much would really be doing the image a disservice ...

(sigh) - a nice vague double negative ... I actually said "To smooth such an edge to get rid of the jaggies would do it a disservice."

but I do like the look of his image better, so maybe you are right.

No "maybe" about it, IMNSHO. :-D

I think maybe the edge you picked is not the best example of a sharp edge Ted, since the bark of that tree might be a bit too bumpy to make a sharp edge. I'm not able to view the image on a big screen right now though, so I guess I could be wrong.

Feel free to pick out a better example Scott. You should not need a big screen to do that, I found my edge with a mere 24" ...

Wow! That's a big one Ted. Yours is SO BIG!



I'm at home, viewing on my 13" MacBook Air now. I was using a little smartphone screen before, which is only like 5" or something like that.

--
Scott Barton Kennelly

Big Print Photos - Photography by Scott Barton Kennelly

Landscape and nature photography, photographic prints by Scott Barton Kennelly, an Australian photographer living in south Florida.

www.bigprintphotos.com

 

[bizi clop]

I made new ones for the DP2M and DP3M, see https://www.dropbox.com/sh/vd601dndp465oq9/AACR7B1D91oiDu3ihvTkTbgKa?dl=0 , would you be so kind and evaluate those too?

I simply went outside on an overcast day, put a white cardboard on the ground and took the photos from about 50 cm above, taking care not to cast a shadow on the cardboard of course. This may not be perfectly homogeneous, but I think it's ok.

I need to confess, it seems that I made a stupid mistake with the old sets, misinterpreting the functionality of LIghtroom's flat-fielding plugin: I applied it to the full set of 32 images (16 test images, 16 reference images) at once, assuming that the plugin would cleverly pair the images shot at the same aperture. Turns out it does not, instead apparently it averages the reference shots and applies the result to all test images - hence the same opcode for all apertures. My bad, embarassing ...

This time I took care to take just the two images at a time, and repeated 16x.
Still surprising that the averaged opcodes do a decent job, at least it was not immediately visible to me that something is wrong ...

Hi, here are they, now in greyscale: left:red/bottom layer, middle:green, right:blue/top layer.

What happened to that poor DP2M? And where was it focused? Is that some dirt/damage on the image stabiliser? That can't be it…

DP1M

DP2M diagnosis: https://en.wikipedia.org/wiki/Floater

DP3M

For enhancing contrast I used the function (tanh(20x-20)+1)/2 for all images which look like this: https://i.ibb.co/bvzz053/tanh.png

 

[xpatUSA]

Here's my result from X3F Wrapper, DNG directly processed in Camera Raw. Manually white balanced on the snow and using a profile home-made with DNG Profile Editor.


DNG from X3F Wrapper 0.56 Beta 3, developed in ACR

Georg, I like your version more, but it does seem to be sharpened more, which causes jaggies on the edge of the flag on that lamp post just right of middle.

But they are not true "jaggies" - 800% pixel-peeping reveals that the edges in that flag are properly anti-aliased, just like Lumo's "perfect pixel" (1.2.2) here:

http://www.falklumo.com/lumolabs/articles/sharpness/index.html

See his "hard pixel" (1.2.1) for true jaggies.

Interesting Ted. Thanks.

Here is what I'm talking about: <examples>

Thanks for the examples but I had already down-loaded Georg's image and viewed it at 800%.

As you can probably tell, you need to not only view at full-size, but then you need to zoom in, so you can see the "jaggies" that I saw, which maybe some people would call stair-stepping. I think the sharpening enhances them basically every time, which is one reason I don't like sharpening.

A matter of terminology, I guess, Scott.

Lumo's hard jaggies or stair-stepping:



Lumo's aliased jaggies or stair-stepping:



The point here is that a perfectly anti-aliased edge is correct and is the best you can get with a digital image. The edges in Georg's flag rendition look much more like the second than they do the first. To smooth such an edge to "get rid of the jaggies" would do it a disservice.

Well Ted, certainly Georg's version of the image looks "sharper" (probably because of more sharpening done in post-processing).

Whenever Scott starts a response with "Well," there's often a rebuttal to follow. Scott continues to say "sharper" and "more sharpening" in spite of my claim of it being about right.

Time to bring in the big guns - QuickMTF and Falk Lumo, whose paper is being studiously ignored :-|

I found an edge close enough to the flag to be significant. Here is living proof that Georg's image is SLIGHTLY SOFT:



Lumo tells us that a perfect sharp edge response is 1.26px rise from 10-90%; Georg's edge above has a 2.70px rise which makes it "OK" but a bit SOFT.

I'm not so sure that NOT sharpening so much would really be doing the image a disservice ...

(sigh) - a nice vague double negative ... I actually said "To smooth such an edge to get rid of the jaggies would do it a disservice."

but I do like the look of his image better, so maybe you are right.

No "maybe" about it, IMNSHO. :-D

I think maybe the edge you picked is not the best example of a sharp edge Ted, since the bark of that tree might be a bit too bumpy to make a sharp edge. I'm not able to view the image on a big screen right now though, so I guess I could be wrong.

Feel free to pick out a better example Scott. You should not need a big screen to do that, I found my edge with a mere 24" ...

Wow! That's a big one Ted. Yours is SO BIG!



I'm at home, viewing on my 13" MacBook Air now. I was using a little smartphone screen before, which is only like 5" or something like that.

That is no excuse for not being willing to identify edges, Scott.

With your continuing comebacks, it is becoming obvious that understanding the technicalities of sharpness is not desired. So good luck with your own methods and opinions.

I'll leave the Last Word to you. :-|

--
WYSINWIG: what you see is not what I got.

 

[xpatUSA]

[No message]

 

[bizi clop]

Thank you!

 

[Scottelly]

Here's my result from X3F Wrapper, DNG directly processed in Camera Raw. Manually white balanced on the snow and using a profile home-made with DNG Profile Editor.


DNG from X3F Wrapper 0.56 Beta 3, developed in ACR

Georg, I like your version more, but it does seem to be sharpened more, which causes jaggies on the edge of the flag on that lamp post just right of middle.

But they are not true "jaggies" - 800% pixel-peeping reveals that the edges in that flag are properly anti-aliased, just like Lumo's "perfect pixel" (1.2.2) here:

http://www.falklumo.com/lumolabs/articles/sharpness/index.html

See his "hard pixel" (1.2.1) for true jaggies.

Interesting Ted. Thanks.

Here is what I'm talking about: <examples>

Thanks for the examples but I had already down-loaded Georg's image and viewed it at 800%.

As you can probably tell, you need to not only view at full-size, but then you need to zoom in, so you can see the "jaggies" that I saw, which maybe some people would call stair-stepping. I think the sharpening enhances them basically every time, which is one reason I don't like sharpening.

A matter of terminology, I guess, Scott.

Lumo's hard jaggies or stair-stepping:



Lumo's aliased jaggies or stair-stepping:



The point here is that a perfectly anti-aliased edge is correct and is the best you can get with a digital image. The edges in Georg's flag rendition look much more like the second than they do the first. To smooth such an edge to "get rid of the jaggies" would do it a disservice.

Well Ted, certainly Georg's version of the image looks "sharper" (probably because of more sharpening done in post-processing).

Whenever Scott starts a response with "Well," there's often a rebuttal to follow. Scott continues to say "sharper" and "more sharpening" in spite of my claim of it being about right.

Time to bring in the big guns - QuickMTF and Falk Lumo, whose paper is being studiously ignored :-|

I found an edge close enough to the flag to be significant. Here is living proof that Georg's image is SLIGHTLY SOFT:



Lumo tells us that a perfect sharp edge response is 1.26px rise from 10-90%; Georg's edge above has a 2.70px rise which makes it "OK" but a bit SOFT.

I'm not so sure that NOT sharpening so much would really be doing the image a disservice ...

(sigh) - a nice vague double negative ... I actually said "To smooth such an edge to get rid of the jaggies would do it a disservice."

but I do like the look of his image better, so maybe you are right.

No "maybe" about it, IMNSHO. :-D

I think maybe the edge you picked is not the best example of a sharp edge Ted, since the bark of that tree might be a bit too bumpy to make a sharp edge. I'm not able to view the image on a big screen right now though, so I guess I could be wrong.

Feel free to pick out a better example Scott. You should not need a big screen to do that, I found my edge with a mere 24" ...

Wow! That's a big one Ted. Yours is SO BIG!



I'm at home, viewing on my 13" MacBook Air now. I was using a little smartphone screen before, which is only like 5" or something like that.

That is no excuse for not being willing to identify edges, Scott.

With your continuing comebacks, it is becoming obvious that understanding the technicalities of sharpness is not desired. So good luck with your own methods and opinions.

I'll leave the Last Word to you. :-|

I took another look Ted, and the only easily-discernible sharp edge I see is almost vertical, and I don't think that will work for you, will it?




See the sharp edge on the right side of the base of that light post?



--
Scott Barton Kennelly

Big Print Photos - Photography by Scott Barton Kennelly

Landscape and nature photography, photographic prints by Scott Barton Kennelly, an Australian photographer living in south Florida.

www.bigprintphotos.com

 

[geni1105]

I made new ones for the DP2M and DP3M, see https://www.dropbox.com/sh/vd601dndp465oq9/AACR7B1D91oiDu3ihvTkTbgKa?dl=0 , would you be so kind and evaluate those too?

I simply went outside on an overcast day, put a white cardboard on the ground and took the photos from about 50 cm above, taking care not to cast a shadow on the cardboard of course. This may not be perfectly homogeneous, but I think it's ok.

I need to confess, it seems that I made a stupid mistake with the old sets, misinterpreting the functionality of LIghtroom's flat-fielding plugin: I applied it to the full set of 32 images (16 test images, 16 reference images) at once, assuming that the plugin would cleverly pair the images shot at the same aperture. Turns out it does not, instead apparently it averages the reference shots and applies the result to all test images - hence the same opcode for all apertures. My bad, embarassing ...

This time I took care to take just the two images at a time, and repeated 16x.
Still surprising that the averaged opcodes do a decent job, at least it was not immediately visible to me that something is wrong ...

Hi, here are they, now in greyscale: left:red/bottom layer, middle:green, right:blue/top layer.

What happened to that poor DP2M? And where was it focused? Is that some dirt/damage on the image stabiliser? That can't be it…

DP1M

DP2M diagnosis: https://en.wikipedia.org/wiki/Floater

DP3M

For enhancing contrast I used the function (tanh(20x-20)+1)/2 for all images which look like this: https://i.ibb.co/bvzz053/tanh.png

I checked on the DP2M: it is due to a very faint blotch on the grey cardboard, barely visible. It is not visible at all at the reference image, it shows up only if I enhance contrast to the extreme. For all practical purposes, it should not matter.

 

[dlj]

Newer flavours of Affinity Photo (1.88+) do fairly good jobs for Merrill images, when they're nicely lit. They're not using default LibRaw for Foveons anymore. However, the processing method they're using now adds noise to significantly underexposed regions for the F13 sensors. It did not before (it actually handled them quite well before). Default colour handling is now pretty decent across about 70% of Foveon images, and now, it will even additively process the 'RGB' into an often useful greyscale, if you had that set in the exposure metadata. The processing environment does a great job of caching regions and using GPUs, such that with a moderate GPU, the overall handling is factors faster than SPP. I still need SPP for a chunk of my images, but I can do a lot more quickly with most images nowadays, so its useful. I normally prefer only global edits, but when I have to hold my nose and 'go photoshoppy', it has a really potent toolkit. It also does all kinds of stacking a little easier than other tools. It occasionally requires baking in destructive changes for complex workflows, but the original RAWs are always fine - you just may have to backtrack a fair bit to fix something, and all those backtracks require an awful lot more disk space than I would imagine, but its helpful when necessary. The workflow mindset and vernacular is a fair bit unusual too (like Adobe's uniqueness, but in its own way): in that regard, it's a bit of a learning curve, like Silkypix was, but this will open almost everything, and work with HDR and high colour and wider gamuts better.

 

[JAL2099]

Thank you for this. Getting 0.56 Beta 3 made all the difference.

Like commented by sambaba above, for Merrills do not use X3F Wrapper 0.57 but use 0.56 Beta 3, to be found under https://github.com/Kalpanika/x3f_wrapper/releases.
By mistake this version lacks the f/5.0 opcode for the DP3M (file DP3M_FF_DNG_Opcodelist3_5.0), but you can easily copy it over from 0.57 in case you have a DP3M.

Regards, Georg