Hi folks, I've been working with a few photographers and engineers to organise a big camera test (as part of a retest for the IQ180/8x10 article featured recently). We're comparing the results from an IQ180 on with Alpa WA/Digaron 40, Cambo/Digaron 35 & Phase 645/Digital 45mm shot in a studio (without flash but with daylight lighting booth). We had photographers who use both large format and MFDB's with us including a Phase One distributor to make sure the photographs were taken correctly. For the 8x10 and 4x5 shots I am happy with post processing these having scanned them on a Howtek 4500 (and checked the scans on a microscope and against a colleagues ICG 800dpi scanner) but I am interested in opinions on the optimal post processing for these files.
Hi Tim,
I'm already looking forward to the results, as it seems to be a well managed and executed process done by openminded people who know what they're doing.
When I'm comparing the 80mp files against the 8x10 4000dpi scans (40,000 x 32,000) should I up rez them and how and also in terms of sharpening etc, do I let Capture One handle everything on auto (probably not) or do I just get a raw capture and use Photokit? etc, etc. We will be printing them as well but they will also obviously be displayed on screen for people to assess as most people won't be able to get to the 'show' we're hoping to give.
There are IMHO 2 aspects to a good film scan. The first is to extract all resolution, and second to minimize the distraction from graininess, and there is a trade-off between those parameters. In practice it means having to scan the film with something like 6000+ PPI by a competent operator who can balance the scanning aperture size with the grain/dye cloud structure. Smaller scans will perhaps not extract all detail, and increase the grain-aliasing effect.
It would therefore be useful to separately determine (and thus demonstrate) what the limiting resolution of a chosen system approach is. With system approach I mean camera system, film, and scanner, their combined MTF in understandable terms. There is a number of methods that could be used to quantify a few simple parameters that everybody can understand, complemented by visual confirmation.
I'm hinting at an Edge Spread Function (ESF) such as reported by the Imatest software as
Edge response (1/3rd down the page), and based on a simple slanted edge that you can print yourself. There is also a slanted edge in my target which spawned a discussion
here on LuLa, and it also easily allows to confirm correct focus. There is a number of LuLa forum contributors who could assist in producing that part of the data, myself included. The method also allows to detect sharpening artifacts such as halo. Since it quantifies a phenomenon in the spatial domain (pixels) it is not difficult to imagine such an edge even without knowledge of how to read MTF curves.
I think resampling would be quite acceptable for all viewers if the up/down sampling operation results in similar looking sharpness (the edge acutance and artifacts balance) if the systems were equally sharp. The only difference that then remains is the actual resolution differences, decoupled from the resampling method. So the resampling method should not introduce a new bias, but that may prove to be trickier than anticipated (e.g. Photoshop is pretty poor at downsampling without introducing artifacts).
The most neutral comparison is therefore to offer both, an upsample of the smaller sensor to the film scan PPI (the only variable introduced is the choice of upsampling algorithm for one of the sources, more about that later), and a downsample to the smaller output size (with the only variable introduced being the downsample algorithm used for the other source). That allows to only have a single variable added to one element of the mix, but both sides of the user spectrum (large output vs smaller output) can draw their own conclusions.
I'm basically trying to avoid people saying "well you should have used Thromboid zero point resharpening with the inconvoluted zeta prime transform four paraboids below sigma 6 criteria" (obviously people may not use these words exactly but I hope you get the idea).
Exactly. Nobody can relate their own workflow to such a theoretically potential quality level, especially when it's only achievable under restrictive laboratory conditions, and depends on the phase of the moon. We all marvel at what rocket science brings, but we if we cannot afford or use it, it's only of theoretical interest.
However, that doesn't mean that lowest common denominator approaches are very informative either, because they introduce distracting artifacts. One could, as reasonable compromise, use resampling methods that are industry proven (e.g. such as implemented in ImageMagick). That will also avoid having to use commercial proprietary resampling algorithms, which may cut some corners for speed, or depends on other settings, while hoping to hide the shortcomings for the uneducated public. The algorithms used by ImageMagick are publicly available and documented, and since the software is free, tests can be independently duplicated with the same results.
We've also shot a scenic to get an idea of colour and real world use (it was a windy day on a cliff edge, just right to test the "you can't shoot 8x10 on a windy day" theory) and the P45+ (which we shot on a Linhof Techno with the 70mm Digaron WA) has a very different colour to the IQ180 (shot with an 80mm on a Phase 645 body). How would you normalise the colour? The IQ180 had a lot better colour in my opinion and tallied a lot closer with the Portra 400 and had a flavour of Velvia in there as well (at least you could get it looking at bit like Velvia which is near impossible with the P45+ as far as I could tell).
Color rendering and tonality differences are another hornet's nest. However, I do wonder how distractive the differences (after white balancing) are? You've seen them, and that perhaps raised the question. Do they appeal to someone's esthetic preference so much as to impede the ability to look past that quality and judge the other qualities?
Cheers,
Bart