This content has been marked as final. Show 6 replies
I have no particular interest in posting to a thread elsewhere...but from reading what was written, there are some gaps in the knowledge of the OP on that thread. I suggest reading ABOUT CAMERA RAW 4.1 which describes what the parameters actually do.
>Please enable sharpen-enabled views at less than 100%
Why, enabling sharpening previews for under 100% will be inaccurate. Why would you want inaccurate?
> I think it may be time to consider a radius of less than 0.5
Testing showed it wasn't needed...so if you want it, you'll have to show why it's needed.
>This may create more complexity than it's worth but you might consider adding a "highlight/shadow protection" slider that would optionally remove the halo dampening
Again, why? You can already ruin images easily enough, but there is no benefit to driving textural highlights to clipping and there's rarely a benefit to sharpening the heck out of deep shadow noise. Both of which can easily be done in Photoshop after opening. Camera Raw's current sharpening is designed specifically to address capture sharpening, not sharpening for effect.
> I have no particular interest in posting to a thread elsewhere...
No problem. I didn't expect anybody here to have to go to the other thread to post a response. Here is fine.
> I suggest reading ABOUT CAMERA RAW 4.1
Actually, I read your article right after you published it but I hadn't checked it recently. Regarding gaps in my ACR sharpening knowledge, I admitted to some of those in the post - particularly exactly HOW ACR does some of the things it does and also the behavior of the Detail slider. I notice that at the time you wrote the article you linked, you admitted not understanding all of the internals of it either :). I find the "halo dampening" of the Detail Slider to really be secondary to the way it emphasizes or de-emphasizes soft edges and fine detail. It also appears to me that ACR is always doing some level of "blend-if"-like protection of highlights and shadows, regardless of how the detail slider is set. I plan on picking up your book while I am back in the States over Christmas so maybe I will find some additional info there.
> Why, enabling sharpening previews for under 100% will be inaccurate. Why would you want inaccurate?
I am sure my reasons won't be convincing for you but here they are:
* So I have a constant visible reminder that I have sharpening engaged (instead of waiting until I have it imported into PS)
* So ACR's behavior is consistent with PS, which does provide the sharpening views at less than 100%
* So I have an inaccurate but better than nothing way of judging overall sharpness of the image. In my mind, I would equate it to something like Soft Proofing. I wouldn't call Soft Proofing "accurate" either but I can understand how some people, with a little effort and practice, can take it to the point where if they see "Y" in soft proofing, they are fairly confident that they will get something close to "X" in the final print. It may be innacurate but maybe it can be innaccurate in a sufficiently consistent way that it is still useful.
> Testing showed it wasn't needed...so if you want it, you'll have to show why it's needed.
I'll take your word for it. It just seemed that based on Mr. Fraser's observations that more mega-pixels = smaller radius that 0.5 might start being a little large and might limit detail extraction for cameras like the 5D, 1ds3, etc.. There is a lot of debate right now around the strength of the AA filter on the 1Ds3 and that is what got me thinking about it.
> Again, why? You can already ruin images easily enough . . .
Yep. I basically agree with you. That is why I had the lukewarm qualifier there. Mainly I was thinking that it could provide "training wheels" for people coming from DPP. It also might help reduce the jaggies on straight line diagonals that I thought might be at least partly due to the dampening of sharpening in the highlights and shadows (like Mr. Fraser's blend-if technique). Not really compelling reasons to add additional complexity but that was what was going through my mind.
Anyway, thanks for responding Jeff.
>Why, enabling sharpening previews for under 100% will be inaccurate. Why would you want inaccurate?
It could be inaccurate but how many times less inaccurate than no sharpening version?
Again ... http://www.adobeforums.com/webx/.3c034372/3 :-)
>I'll take your word for it. It just seemed that based on Mr. Fraser's observations that more mega-pixels = smaller radius that 0.5 might start being a little large and might limit detail extraction for cameras like the 5D, 1ds3, etc
The radius is in pixel units, so the number of megapixels of the camera does not matter. In fact, in general, the higher the megapixels, the more you are limited in resolution by diffraction, motion smear, etc. So in general, for higher MP cameras you would need higher numbers for the sharpening radius instead of lower. Apart from that, even if everything is ideal, there is no way a Bayer-type camera can capture detail below 1 pixel anyway, let alone 0.5 pixels.
> So in general, for higher MP cameras you would need higher numbers for the sharpening radius instead of lower.
On the surface that makes sense but I think it is more complex than that and somewhat counter-intuitive. Bruce Fraser, in his image sharpening book, recommended capture sharpening radii of 1 for <= 4 mega-pixel sensors down to 0.4 for >= 11 mega-pixels. He certainly studied the topic and experimented with images much more than I have. Unfortunately, I have not found in the book nor in other articles on the web where the exact mechanics are explained. He does say that a too large radius will start to block fine detail and too small of one will lower the sharpening dramatically and not reveal any additional detail.
> Apart from that, even if everything is ideal, there is no way a Bayer-type camera can capture detail below 1 pixel anyway, let alone 0.5 pixels.
Again, sounds logical but, again, I think it is trickier than that. Remember that in operations of this type, the overall effect is determined by the interaction of multiple pixels - not just one pixel on its own. I can't explain the math but think about a detail speck that has been "blurred" by the AA-filter/bayer process and now has impacted the information on a 2x2 grid of pixels in the image. It seems to me that a 0.5 radius sharpening action could reveal that speck better than a 1 radius given the interaction of the sharpening of the 4 pixels. I am probably not explaining it correctly but I suspect it is something along those lines.
>1 for <= 4 mega-pixel sensors down to 0.4 for >= 11 mega-pixels
I know, but this is completely counter to what I have found. In general, the lower MP cameras I use benefit from 0.5 sharpening radii, but the higher MP cameras generally need 1 pixel or so when used with the same sharp lenses. Note that I don't have a Full frame sensor camera. It is likely that those would behave similar to the 6 MP cameras as the physical size of the pixel sensor is more or less the same in the FF >10MP cameras as in a DX 6MP camera. So it is possible that he was talking about a hidden variable - sensor size. If the 4MP cameras he was thinking about have very small sensors and the 10MP cameras have bigger ones than indeed the 10MP ones might need smaller sharpening radii, simply because because of the lens blur not being as limiting in the big sensor case as in the small sensor case. However, if the sensor is the same size, there is no way a higher MP sensor needs a smaller sharpening radius.
>t seems to me that a 0.5 radius sharpening action could reveal that speck better than a 1 radius given the interaction of the sharpening of the 4 pixels
This doesn't make much sense to me. The speck is impacting a 2x2 grid! This means the sharpening should be about 1-2 pixels not 0.5! If anything, the physical resolution limit of a bayer sensor is about half of what they state (so a 10 MP camera can only capture about 5 MP real info) indeed because of the Bayer mosaic that is employed. It is a little higher if you only consider green light, since there are twice as many green sensors as there are red or blue ones, but not much more. If you think linearly, this means that the maximum detail you can capture using a Bayer sensor is about 1.4 pixels wide regardless of the MP size of the camera. It is a little smaller for green light and a little larger for red and blue. This is before you start accounting for the AA filter, lens/diffraction blur and chromatic aberrations. The latter two become worse at higher MPs, making higher MP cameras need larger sharpening radii for capture sharpening.