It is Finally Here, 907x 100c

[Ray Harrison]

Is it even worth patenting? The result is just an in-camera equivalent of what you can get via PS's 'stacked images > smart object > stack mode > mean (or median)' workflow - people have been doing it that way in PS for years, and I believe even many smart phones offer it as well.

I don’t know whether they’ve patented this over at Phase, but what I like about their approach is they can do this over hundreds+ of images and spit out a raw file at the end that is written in normal raw write time. So it’s like Matt is saying, that they’re processing quickly with each capture, but the additional feature of having a raw file at the end is very nice. Piping several hundred images through PS isn’t something I’d want to do too often (personally).

 

[leejo]

I've been completely satisfied that it does everything I wanted to do with it when I bought it very nicely indeed, and I suspect the 100 mpixel version does the same. I just don't know that it's actually worthwhile for me to spend another $8000+ to get that doubling of pixel resolution (about 35% linear resolution gain; it would take about 200 Mpixels to double the linear resolution of the CFVII 50c), I don't really use it enough for that or need that additional resolution edge.

I've found that the most compelling reason to have the resolution is having cropping options that removes the need for longer lenses, which I can't carry around (tour and snowboard) with me on the pistes anyway. If you're not cropping then I don't think there's a reason to upgrade. I bought the GFX 100S, have stuck a 50mm on the front, and found I need nothing else so far. Wider? Stitch. Longer? Crop. I've been printing a few large panoramic images (1.5m on the long edge) from significant crops and they look great.

It should be noted that the cropping option is why I'm concerned about things like IBIS and such. I was cropping with the CFV II 50c as well, but more resolution gives me more options and doesn't require any upscaling. In essence it's a return to the original medium and large format film workflow: frame loose, keep your crop options open.

I'm considering selling my entire Xpan setup at the end of this winter season due to the improvement in workflow (which would nicely fund the 30mm T/S at current prices...).

 

[f8orbust]

I don’t know whether they’ve patented this over at Phase, but what I like about their approach is they can do this over hundreds+ of images and spit out a raw file at the end that is written in normal raw write time. So it’s like Matt is saying, that they’re processing quickly with each capture, but the additional feature of having a raw file at the end is very nice. Piping several hundred images through PS isn’t something I’d want to do too often (personally).

Jim Kasson has a small script for Matlab (posted in 2013) which processes the images (in double-precision floating point) extremely quickly. Surprisingly simple to do - wouldn't surprise me if P1's implementation is almost identical. It can handle hundreds of images too, though I'm not sure why you would actually take hundreds of images (e.g. you can reduce noise by 50% with only four captures).

 

[mristuccia]

Here is the "running average" formula:

m0 = 0
mn = mn-1 + ((vn - mn-1) / n)

where:
n is the current iteration (current shot number)
mn is the running average of the pixel at iteration n
vn is the pixel value at iteration n

Each subsequent step is only based on the current pixel value (aka the last shot), the current iteration number and the average cumulated in the preceding step (aka the buffer Matt was referring to). You can stop this process at whatever iteration (amount of images) you want, without knowing the amount of images in anticipation. The buffer will contain the averaged result to be saved into the output file.

I don't see any reason to patent a software/firmware based on such formula which is known and public.

 

[MGrayson]

Here is the "running average" formula:

m0 = 0
mn = mn-1 + ((vn - mn-1) / n)

where:
n is the current iteration (current shot number)
mn is the running average of the pixel at iteration n
vn is the pixel value at iteration n

Each subsequent step is only based on the current pixel value (aka the last shot), the current iteration number and the average cumulated in the preceding step (aka the buffer Matt was referring to). You can stop this process at whatever iteration (amount of images) you want, without knowing the amount of images in anticipation. The buffer will contain the averaged result to be saved into the output file.

I don't see any reason to patent a software/firmware based on such formula which is known and public.

Or simply add them all up in a buffer with a 8 or 16 extra bits and divide at the end. (Not my idea - I did it your way.)

 

[mristuccia]

Or simply add them all up in a buffer with a 8 or 16 extra bits and divide at the end. (Not my idea - I did it your way.)

I'd prefer to do it gradually, just to be sure we do not overflow the buffer.

 

[MGrayson]

I'd prefer to do it gradually, just to be sure we do not overflow the buffer.

Me, too, but that (probably) means floating point operations and a lot of multiplications. I don't know where the data/computation bottlenecks are in this process, but whatever you do, you do it 10^8 times per shot!

A limit of 2^8 images may be too small, but 2^16 should be enough even for @Ed Hurst .

 

[Ray Harrison]

Jim Kasson has a small script for Matlab (posted in 2013) which processes the images (in double-precision floating point) extremely quickly. Surprisingly simple to do - wouldn't surprise me if P1's implementation is almost identical. It can handle hundreds of images too, though I'm not sure why you would actually take hundreds of images (e.g. you can reduce noise by 50% with only four captures).

Sure. It’s not just a noise reducer, though, right? It is another way to build a long-exposure-like image. And for sure, there are plenty of ways to get to an end result like that and I certainly have no vested interest in convincing anyone that there’s a “right way” or a “best tech”. It’s awesome that Jim’s got a script. Knowing his work, I’m sure it’s terrific.

 

[davidsuchoff]

This is unrelated to the current conversation, but can anyone speak to the battery life on the 907x 100c and how it compares to the prior model? One of the knocks on the 50c was lackluster battery life.

 

[TechTalk]

by extension one "could" be able to apply the HNCS profile within Phocus to non Hasselblad files by cracking the metadata so it looks like from a Hassy digital back – IF – there are not more fancy safeguards in place within Phocus (ie the gain file for the sensor normalization step).

At least a Leica S rep told me thay were able to open S files in Phocus somehow and apply the colour science. Maybe with an old version of Phocus it could work ... if it doesnt with the current one.

If that works one can also use Phocus as a fancy in-between LUT tool. And then you could even create a profile from the before and after and save it as an input ICC profile in C1, lol.

There's more to Hasselblad's color system than a profile. It begins with producing individual sensor calibration data — which is very extensive to cover an extremely wide-ranging set of exposure conditions — that is stored in the camera and applied to the raw image data. It ends with Phocus which has its own unique system for reproducing color from a variety of lighting conditions. What makes the Hasselblad Natural Color Solution (HNCS) so well liked (in my personal opinion) is not so much that the color is "good" or "natural" under the right conditions, but that color is reliably and consistently reproduced in a way which is pleasing for a wide range of images captured under various conditions.

The deepest look into how Phocus works I've seen was from Anders Torger (who created Lumariver Profile Designer software) which he wrote about here. His conclusion was: "...Phocus will automatically select illuminant matrix and chroma correction LUT based on the white balance you set in Phocus, and blend between them for intermediate values, that is the same way dual illuminant DCP works in Lightroom. The difference from DCP is that Phocus profiles have more illuminants, and is more sophisticated in that aspect, but have a bit simpler LUT. On the other hand will the simpler LUT [will] make sure that they're no nasty non-linearities which you can have in (bad) DCPs."

"In all I'm quite impressed with the color model in Phocus, it's simple towards the user and advanced under the surface."

Further down in that same thread, Anders Torger mentions Leica and Phocus: "Looking at naming of files they seem to have this type of profile for some Leica system, but otherwise only their own cameras." The Leica system which is (or was) supported in both Phocus and the earlier FlexColor software is the unique Leica Digital-Modul-R (DMR). This was a collaboration between Leica (camera body), Kodak (image sensor), and Imacon/Hasselblad who designed and manufactured the back for Leica. The project started with Imacon in 2003 a year before the acquisition which folded Imacon into Hasselblad. The Leica DMR first shipped in 2005 manufactured by Hasselblad in Denmark post-merger.

As many are aware, Hasselblad and Adobe engineers collaborated to make Hasselblad raw files processed in Adobe software look as close to images from Phocus as possible — at least under most exposure and lighting situations. Hasselblad then included a full licensed version of Lightroom with each new camera, along with Phocus, for some time afterward. The purpose for Hasselblad doing this was to give customers a choice and, most importantly, to lower any barrier to purchasing a camera for customers already experienced and comfortable using Lightroom or Adobe Camera Raw. It appears they succeeded. Many Hasselblad owners use Lightroom for processing their images and are happy with the results.

 

[tcdeveau]

That is correct.

Phase One owns that technology and how they developed it, so I do not expect any of these other medium format manufactures to offer it. Wish they would, it is two great features. It does require a lot of power and processing internal to the back.

Why do you think they own the technology? I could not find a patent on it. No patent, no excluding others from using it (I.e., no ownership) It’s possible someone else has one and they have exclusive license or something.

Plus Sony offered frame averaging in their smooth reflections app years before phase one, which would make it difficult for P1 to patent.

 

[tcdeveau]

Anyone have the patent? I actually have not found any for frame averaging (so far).

They do not have a patent on it, and least not under the name “Phase One A/S.” I posted somewhere recently about what what I found in my latest patent search.

Patent rights are also jurisdictional so they’d have to have one in every jurisdiction in which they wanted to exclude others from using it.

 

[tcdeveau]

Even if they have a patent on it, patents expire and then the technology that was patented and disclosed is freely available.

Yes, but current law in the us is, as long a maintenance fees are paid, patents in the US expire 20 years after the earliest filing, so it takes time (photography doesn’t pay the bills so I decided to be a patent atty instead).

 

[Paul Spinnler]

There's more to Hasselblad's color system than a profile. It begins with producing individual sensor calibration data — which is very extensive to cover an extremely wide-ranging set exposure conditions — that is stored in the camera and applied to the raw image data. It ends with Phocus which has its own unique system for reproducing color from a variety of lighting conditions. What makes the Hasselblad Natural Color Solution (HNCS) so well liked (in my personal opinion) is not so much that the color is "good" or "natural" under the right conditions, but that color is reliably and consistently reproduced in a way which is pleasing for a wide range of images captured under various conditions.

The deepest look into how Phocus works I've seen was from Anders Torger (who created Lumariver Profile Designer software) which he wrote about here. His conclusion was: "...Phocus will automatically select illuminant matrix and chroma correction LUT based on the white balance you set in Phocus, and blend between them for intermediate values, that is the same way dual illuminant DCP works in Lightroom. The difference from DCP is that Phocus profiles have more illuminants, and is more sophisticated in that aspect, but have a bit simpler LUT. On the other hand will the simpler LUT [will] make sure that they're no nasty non-linearities which you can have in (bad) DCPs."

"In all I'm quite impressed with the color model in Phocus, it's simple towards the user and advanced under the surface."

Further down in that same thread, Anders Torger mentions Leica and Phocus: "Looking at naming of files they seem to have this type of profile for some Leica system, but otherwise only their own cameras." The Leica system which is (or was) supported in both Phocus and the earlier FlexColor software is the unique Leica Digital-Modul-R (DMR). This was a collaboration between Leica (camera body), Kodak (image sensor), and Imacon/Hasselblad who designed and manufactured the back for Leica. The project started with Imacon in 2003 a year before the acquisition which folded Imacon into Hasselblad. The Leica DMR first shipped in 2005 manufactured by Hasselblad in Denmark post-merger.

As many are aware, Hasselblad and Adobe engineers collaborated to make Hasselblad raw files processed in Adobe software look as close to images from Phocus as possible — at least under most exposure and lighting situations. Hasselblad then included a full licensed version of Lightroom with each new camera, along with Phocus, for some time afterward. The purpose for Hasselblad doing this was to give customers a choice and, most importantly, to lower any barrier to purchasing a camera for customers already experienced and comfortable using Lightroom or Adobe Camera Raw. It appears they succeeded. Many Hasselblad owners use Lightroom for processing their images and are happy with the results.

You are repeating what I've often mentioned. And you are misrepresenting what I said.

On multiple occasions I referred back to Toger's post where he explained that Phocus interpolates between illuminants as well as the normalisation step based on the sensor. After all, he's the author of Lumariver, so he understands a thing or two about colour profiles. I even have it installed. That's clear. Leica S also has a normalization step via a gain file I've been told by Leica and in the past it was at one point apparently possible to open S files within Phocus with some tricks. This is literally what I was told a few years back by a Leica S rep, so I am not sure uf this is still possible and whether one needs to edit metadata to do so.

When mentioning that it is a fancy LUT it is actually that – there's some more logic to how it is essentially applied in the sense that it modulates the application depending on the lighting condition and it has a normalization step before so the colours look the same with all cameras – ie kelvin of the scene, but the basis of HNCS is to prioritize natural skin colour rendition and an overall pleasing look based on many feedback loops with photographers during the time it was developed. The end product, with tone curve adjustments on top, is an "analogue, film like" rendition:


As described here and in this webinar:

So these colour transformations are reminiscent of what film does especially in terms of skin tone unification, and the saturation and depth of certain colours.

 

[TechTalk]

You are repeating what I've often mentioned. And you are misrepresenting what I said.

I simply expanded on the subject of Hasselblad's color system, how it works, and its relationship to Leica and Adobe. Please relax, it's not about you.

 

[Paul Spinnler]

I simply expanded on the subject of Hasselblad's color system, how it works, and its relationship to Leica and Adobe. Please relax, it's not about you.

The point is you quote my post and then start with that there's more to that XYZ, implying that I presented something in the wrong way – but actually this is what I had said many times already ...

In any case. HNCS is a fancy, interpolated LUT that is applied on a sensor normalized basis – a very good profile. If you've ever created a profile – it is easy to do one, but very time consuming and difficult to create a great one which works under all conditions and doesn't tear up color ranges, ie does very smooth lookups across the whole colour spectrum. Often times "sloppy profiles" create artifcats in edge cases and I understand that HNCS works perfectly all the time.

This means HNCS is a very robust profile which can be universally used without artifacts, which is an achievement.



In a very good LUT you have very smooth transitions across all colours and brightness levels. This is a good LUT. HNCS would probably be similarly good without big artifacts.

 

[TechTalk]

There's more to it than simply a LUT. That was one part of my post.

Look, if you want me to delete my post I'll do so. I'm just offering additional information in which some may be interested, but it's not critical insight or anything which hasn't been written about before both here and elsewhere by myself and numerous others.

In the future, I'll simply avoid quoting you at all on any topic or in any context.

 

[Paul Spinnler]

You don't need to delete, its fine. You have or had a way of quoting to correct, and less so quoting to add. That's why there's a reaction, I will chill and hope we can all agree how great HNCS is. That's the only feature I wish P1 would look into. It wouldn't be that difficult to mimic, they even have a color professor, not sure he's still there, though. ...

All about color with Niels Knudsen

photography.phaseone.com

I always wondered why P1 seemingly was always focused on neutral colour rendition and never tried to copy the approach of HNCS. This is THE killer feature everyone always refers to with Hasselblad. Leica S also has its own analogue look, but they don't give it a name. I can tell, though, that S files look very different than all other Leica files too.

This being said, sorry for my reaction. Are we good?

 

[TechTalk]

Not a problem.

 

[Paul Spinnler]

Great!