Aham, cool tip. How about sharpening applied in the resize? If I have 3 step sharpen/resize script it means I have to repeat the same procedure each time? It is going to take ages to re-work each shot..though...perhaps choosing 3-5 only from a set will do..
I think that if you make sharpened layer with layer mask on the full size picture (right after import), it will resize together with the rest of the layers. So you could mask the bokeh/sky or other parts that don't need sharpening right at the start. The layer mask doesn't have to be so precise.
By the way, Lightroom (if you use that to export your files) has the "masking" slider in the sharpening tab, which masks all areas that have very little or low contrast edges in them, so that can be quite useful for capture sharpening.
Did you base your script on Denoir's/Samuli's one that they used for their 5Dmk2? And you have a 5D classic, right? One of the first things I noticed when trying out denoir's script is that the regular sharpen before resizing should sometimes be skipped because it affects bokeh a lot and doesn't increase detail that much, or in a bad way.
Also, from my testing I concluded that the 5D files don't need more than a 2-step resize. I didn't see a noticeable improvement with 3 steps or more.
AhamB wrote:
By the way, Lightroom (if you use that to export your files) has the "masking" slider in the sharpening tab, which masks all areas that have very little or low contrast edges in them, so that can be quite useful for capture sharpening.
this is exactly what I am doing in my sharpening script in PS. In the first sharpening step (unsharp mask) I let the dialog pop up and use threshold to control the areas where sharpening would destroy the image. I mostly end up with a value of 3 to 6.
Well, it shows in the actual image (I always have that set to 1:1) and in the preview pane but it is not marked like with a red mask or so (or I have not found a way to show that). But if you zoom to an area where sharp edges are next to OOF background, the threshold slider shows what it does.
I usually wind up in that same 4-5 range, with some subjects that warrant a preceeding USM at 9-12 threshold.
I rarely sharpen with threshold @ 0
Highly variable depending on the subject, but mine usually goes something like this.
375, .8, 9-12
170, .4, 5
22, 55, 3-20 ... usually followed by fade (typically around 70%) and finished off with a slight contrast reduction (-8) to take a little edge off for a bit easier viewing (optional).
Lately, I've working on sharpening individual layers, like the K in CMYK and the L in LAB. It seems that you can push the neutral layers even harder without incurring horrific penalty ... I'll go above 400 and above 1.0 unlike when sharpening RGB. Haven't settled on anything for a 'base recipe' with those just yet ... still learning how all the pieces "play nice" with each other.
@Rusty: So no smart sharpen? I found that nothing brings out the finest detail like smart sharpen with 0.1-0.3 radius and "more accurate" checked (I usually do multiple smart sharpens with strength 30-40). I think repeated smart sharpens achieve something similar to deconvolution sharpening (according to some experiment by Lloyd Chambers IIRC).
Yes, there may be some deconvolution that is occurring, but via the threshold variation it seems to work to achieve a look that I like. I built some B&W test patterns as part of my experimenting to see how the multi-step processes were playing with each other ... fiddling till I got the sharpening / -halo effect that I was please with.
I played with smart sharpen some time back, but for some reason it never really grabbed me ... most likely the "old dog, new tricks" syndrome . I might need to revisit it.
BTW ... I definitely go down to .3 and sometimes .2 (almost never .1). I just threw up those numbers for a 'relative' reference, not so much a literal action. Each image gets its own "recipe" depending on what areas I'm trying to reveal vs. safeguard within the image.
Threshold is nice for safeguarding things like skies and flower petals without the need to try & mask while you sharpen the veins & clouds.
AhamB wrote:
I think that if you make sharpened layer with layer mask on the full size picture (right after import), it will resize together with the rest of the layers. So you could mask the bokeh/sky or other parts that don't need sharpening right at the start. The layer mask doesn't have to be so precise.
By the way, Lightroom (if you use that to export your files) has the "masking" slider in the sharpening tab, which masks all areas that have very little or low contrast edges in them, so that can be quite useful for capture sharpening.
Did you base your script on Denoir's/Samuli's one that they used for their 5Dmk2? And you have a 5D classic, right? One of the first things I noticed when trying out denoir's script is that the regular sharpen before resizing should sometimes be skipped because it affects bokeh a lot and doesn't increase detail that much, or in a bad way.
Also, from my testing I concluded that the 5D files don't need more than a 2-step resize. I didn't see a noticeable improvement with 3 steps or more....Show more →
Yepp, I skip the initial sharpening at all for most lenses, with exception of perhaps WA Zuikos, which I almost never use at all nowdays..
I also decreased the radius and intensity of the sharpening since it resulted in oversharpened shots.
So, since I work in LAB, I would need to convert first, then duplicate layer and use it as a mask? Hmm...possibly.
I skip Lightroom and CaptureOne because I hate the database and workflow management. I don't shoot sessions at the moment and don't want to waste megs on previews and such.. so, I pretty much use CRAW in PS.
Rusty, what is the 110mm like stopped down? I'm looking to find a cheap lens for stitching urban landscapes which will come close or match the Canon 100mm f2/70-200L f4 I used to use for this task now sadly long sold. I have a Mamiya shift adaptor already and the 110mm is very cheap so I'm wondering how sharp/contrasty/flare resistant and how well resolving the lens is at the f16/22 mark compared to a modern 35mm format lens. Is it in the ballpark?
I've never compared them ... I mostly use the 100/2 for low light or AF duty (both of which are rare for me) ... but can put up a shot or two. Not much around here to look at (i.e. urban landscape) ... but are you looking to focus at infinity or something much less?
If you are looking to stitch, I think the M645's with their larger image circle work well for reduced vignetting variation.
Krosavcheg wrote:
I recall somebody recommending staying away from 110mm and 45mm Mamiyas...
Hmmm ... I've got both.
I think it's a matter of what you are expecting from them. I like the M645 glass because of it's smoothness of tonal transitions and neutrality ... that and the larger image circle makes vignetting less of an issue. If you are an a mission for max punch/contrast/microcontrast/etc. ... you may not find them to your liking as much as say a ZE 100/2.
Krosavcheg wrote:
I recall somebody recommending staying away from 110mm and 45mm Mamiyas...
I have both (and both are N variants) and I like both. However, as seems to be the case with Mamiya glass in general, they do suffer a bit from CA wide open which may or may not be a problem. The 110 is particularly nice as a portrait lens and I think this was it's purpose anyway. It seems to have less SA correction so seems to have reasonably nice Bokeh, but is possibly not as sharp/contrasty wide open as a result.
RustyBug wrote:
I've got the Canon 100/2 and the M645 110/2.8 N
I've never compared them ... I mostly use the 100/2 for low light or AF duty (both of which are rare for me) ... but can put up a shot or two. Not much around here to look at (i.e. urban landscape) ... but are you looking to focus at infinity or something much less?
If you are looking to stitch, I think the M645's with their larger image circle work well for reduced vignetting variation.