Lens Design - should it rely on software

[Morgan_Moore]

Some seem to think that the phisical engineering of the blad 28 not to phisically correct abberattions that can be corrected in software is in some way an impure approach

Others would argue that the marriage of hardware and software leads to advantages in price/size/performance - I would be one

Further this issue is clouded by hasselblad not making this lens and the distortion algorimths apertaining to that 'open source'

I would argue that this is for marketing rather than technical reasons

The debate of that is a secondary subject to the primary issue of the 'purity' of digital correction

Debate....

SMM

[jonstewart]

Well put. That just about sums it up. As I indicated in one of my posts, I think my opinion is changing on the matter, but still don't like the 'proprietary' and secret algorithm.

Is that algorithm also built into the camera body firmware, so that you see the 'corrected' version on the LCD screen, and in fact, does it extend to tethered shooting?

(The alternative would be amusing, with the AD etc looking over your shoulder at an uncorrected photo, and the photographer explaining, that's how it's not really going to look... I mean that's what tethered shooting is all about?)

[Graham Mitchell]

Any interpolation will reduce image quality, so a lens which doesn't require correction will outperform one which does, all else being equal.

[josayeruk]

Well put. That just about sums it up. As I indicated in one of my posts, I think my opinion is changing on the matter, but still don't like the 'proprietary' and secret algorithm.

Is that algorithm also built into the camera body firmware, so that you see the 'corrected' version on the LCD screen, and in fact, does it extend to tethered shooting?

(The alternative would be amusing, with the AD etc looking over your shoulder at an uncorrected photo, and the photographer explaining, that's how it's not really going to look... I mean that's what tethered shooting is all about?)
[a href=\"index.php?act=findpost&pid=144888\"][{POST_SNAPBACK}][/a]

I imagine it is secrect for lots of reasons.  I guess the main one is cost - R&D cost.  Would you just give that away?  Not saying yes or no, just asking the question.

If it were licenced would companies be willing to buy the correction details.  Also, would other systems be able to use that data effectively?  Could a Phase Back read lens data without being modified?

The corrections are not shown on the LCD, but I think it would be pretty hard to see the difference anyway as it is not a huge correction... and the LCD is tiny anyway!  

Correction is instantly applied with tethered shooting and it really is instant.  No processing time at all.

Jo S.x

[Morgan_Moore]

I imagine it is secrect for lots of reasons. [a href=\"index.php?act=findpost&pid=144894\"][{POST_SNAPBACK}][/a]

I think it only sectret for one reason - its a marketing lock out

Actualy producing a 'performance matrix' (my expresssion)

Would seem pretty simple if that sort of thing is your job

Photograph a chequerboard lit by RG and B light

Write a 'transform' that alligns all three and makes them chequerboard shaped

repeat the process for a few focus distances and apertures

use some maths to make a full matrix

A bit like PS>filter>liquify meshs in three colours

When shooting record aperture and focus distance and apply

Even if a third party back did not get the ap/distance info then a softwre sliders could be used for 'manual correction'

I would love to see a chipped 28 stuck onto DXO software

Its just a lock out IMO

SMM

[Morgan_Moore]

Any interpolation will reduce image quality, so a lens which doesn't require correction will outperform one which does, all else being equal.
[a href=\"index.php?act=findpost&pid=144892\"][{POST_SNAPBACK}][/a]

Ulitmately yes, true, but it is is what is acceptable on a price/performance/size curve

And actually do we know about even doing this physdically - software might be better !

Now I dont understand chips but I think they record seperate RGB

so a lens splits the RGB like the Pink Floyd album cover and then coatings try to stuff them back together

Maybe just muddling up how the pixel data is mapped could be equally  'pure'

S

[Bernd B.]

have a look at this thread:

http://luminous-landscape.com/forum/index....18169&hl=conrus



Khun_K   
  Jul 15 2007, 10:59 AM
Post #13


members


Group: Members
Posts: 148
Joined: 7-May 06
From: Bangkok, Thailand
Member No.: 17,356



I have the H3D39 system and the 28mm lens. I did mount the lens on a friend's H2 system (but he uses P45+) and the H2's shutter just won't work when the 28mm lens is mounted on. So is my H3D39 back won't work with the H2 (obviously without the power supply) but even tethered the camera just won't work. The 28mm lens is OK for such wide angle lens, but it is not superior than the 35mm or 50mm, except it is wider. Even without the DAC the lens is OK with little distortion and visible chromatic issues. But the lens is just OK, it does not have a strong character, just a superwide, reasonable sharp and good corrected lens. The 28mm is visibly behind the Schneider Digitar 24mm I have for my Alpa, and I will asume it is behind the HR 28mm from Rodenstock.

[josayeruk]

Any interpolation will reduce image quality, so a lens which doesn't require correction will outperform one which does, all else being equal.
[a href=\"index.php?act=findpost&pid=144892\"][{POST_SNAPBACK}][/a]

If you compare the image before / after correction there is no noticeable difference.

[Morgan_Moore]

the H2's shutter just won't work when the 28mm lens is mounted on. So is my H3D39 back won't work with the H2 [a href=\"index.php?act=findpost&pid=144900\"][{POST_SNAPBACK}][/a]

This is just software/firmware lockout is it not - pure market share holding

Thats where chipping the lens would be fun

not that I have the budget or skill

rip apart a 35 and a 28 stuff the electronics from the 35 into the 28 and ...

S


(There is a bloke who cracked contax to make them go on canon (or the other way round)

[ErikKaffehr]

Hi,

I think it may be reasonable to correct distorsion and perhaps chromatic aberration in software. More effort could than be spent on field curvature, spherical aberration, astigmatism and coma.

Best regards

Erik

Some seem to think that the phisical engineering of the blad 28 not to phisically correct abberattions that can be corrected in software is in some way an impure approach

Others would argue that the marriage of hardware and software leads to advantages in price/size/performance - I would be one

Further this issue is clouded by hasselblad not making this lens and the distortion algorimths apertaining to that 'open source'

I would argue that this is for marketing rather than technical reasons

The debate of that is a secondary subject to the primary issue of the 'purity' of digital correction

Debate....

SMM
[a href=\"index.php?act=findpost&pid=144886\"][{POST_SNAPBACK}][/a]

[Dustbak]

I for one applaud the whole concept of correcting known errors and phenomena in software.

We already work with 'gain adjustment', 'custom white' etc. to correct all sorts of other optical defects/weak points.

To use software to leviate part of the work burden from the lens is IMO a good approach.

I think the corrections are really lens specific so I don't see that much benefit of keeping the correction algorythms secret. What can be a downside, other back manufacturers integrating the correction of Hasselblad lenses may mean less back sales but will be offset by more lens sales. I wonder whether Hasselblad has made calculations of both approaches and their results. Would be very interesting.

I see a lot of benefit of sharing it, support via more software suppliers, better solutions. Also the ability to use software corrected lenses with other backs and the H1 & H2 when the software correction is modulized and taken out of the back.

It would hurt more when others could take this data and use it on their own lens designs. I somehow believe that is either not possible or difficult?

[Morgan_Moore]

If you compare the image before / after correction there is no noticeable difference.
[a href=\"index.php?act=findpost&pid=144901\"][{POST_SNAPBACK}][/a]

noticable to your standards

(and most likely mine   )

but that is not the approach of an engineer

S

[pss]

i am not an engeneer, i only take pictures....but there are several analogies that this reminds me of: printing a file from a P45 on a 5 year old 99$ printer, listening to a am mono station through 100000$ speakers and the guy who always says on set : "we can fix that in post production"

you want the best possible start to get the best possible result...if you have crap to start with, all the  software magic in teh world won't give you a good result and do a little extra NOW so you don't have to fix it later when it might be too hard to fix...

i think that all in all software lens correction is great, because no lens is 100% flaw free....never will be...but even with correction it won't be....so the ideal situation is a 95% lens that is brought up to 98% through software....

knowing companies and seeing how hard it must be to actually build 95% lenses it makes much mre sense financially to put a lot of money into software slutions that can be applied to all lenses and simply build lenses at a much cheaper 80% level.....i am not saying that hasselblad is doing that (yet) but i can see that it would make sense for them in the future....

i would really hate to see the next generation of chips going up to 3200 and finding out that this is done purely by applying software....

the moment you push pixels around one way or another you loose information and i would rather keep that to a minimum.....

[Graham Mitchell]

Will the savings be passed on to the consumer? I doubt it. The Hass 28mm is nearly $4K.

[Photomangreg]

Will the savings be passed on to the consumer? I doubt it. The Hass 28mm is nearly $4K.
[a href=\"index.php?act=findpost&pid=144981\"][{POST_SNAPBACK}][/a]

Yeah, but the Mamiya 28mm is well over $5000, and considering the average price of Hassy lenses are twice as much as the Mamiya lenses, I guess they did pass the savings along!

[Morgan_Moore]

if you have crap to start with, all the  software magic in the world won't give you a good result [a href=\"index.php?act=findpost&pid=144971\"][{POST_SNAPBACK}][/a]

I dont see it quite that way

Capturing wide angle image is kind of a strange thing to do rendering them across a flat plane anyway

The best way of displaying a WA image would be to print it on the inside of a ball, stick your head in the ball and have a look around  

So there are always distortions printing wide on a flat image - or capturing it on a flat sensor*

Thats why a man photographed from ontop has big head and small feet - a huge distortion but a visually accepted way of projecting a 3d view onto a flat plane

So I bet that fisheye is a purer form of capture and fisheyelenses can have better LPI and less abberation than a 'correcting' lens given the same quality of engineering

Therefore captureing a fisheye image the captured information may not be 'crap'
just good information presented in a non traditional manner

Using software to then defish the image may therefore render higher res and less abberations

SMM

*the same is true for long lenses but once you are looking a a small enough section of a large enough 'ball' the effect is less dramatic

[jimgolden]

Hasselblad knows itself all too well, they know they'd have to charge at least $12k USD if the Mamiya version is $5500 USD....hell lets make it $15k - call it ultra-megha phocus lead free version - we all know someone will buy it, probably a lot of people!

[jimgolden]

i would really hate to see the next generation of chips going up to 3200 and finding out that this is done purely by applying software....


[a href=\"index.php?act=findpost&pid=144971\"][{POST_SNAPBACK}][/a]

hasselblat already doing that w/ their new Flexcolour or Phocus, older backs are gaining a stop or something...

[stefan marquardt]

So I bet that fisheye is a purer form of capture and fisheyelenses can have better LPI and less abberation than a 'correcting' lens given the same quality of engineering

Therefore captureing a fisheye image the captured information may not be 'crap'
just good information presented in a non traditional manner

Using software to then defish the image may therefore render higher res and less abberations

[a href=\"index.php?act=findpost&pid=145015\"][{POST_SNAPBACK}][/a]

on my quest to find out, if it´s worth it to get the 28mm mamiya, yesterday I bought an old ukrainian 30mm arsat for 80 euros. first tests left me quite suprised how sharp it is. after the image is corrected for CA and defished I get a very undistorted picture. its actualy easier to deal with, than a normal wideangle with a unsymetrical distortion (wave, moustache..). the problem is, after its defished it looses some angle of view - from about 120 down to 100. still slightly more than a 35mm wide angle.  perhaps about the same aov as a 28mm lens?

would i pass the savings - if I would use the arsat instead of the 5000$ mamiya - onto my clients - just like hasselblad - yes course ;-)


stefan

[TechTalk]

Hasselblad knows itself all too well, they know they'd have to charge at least $12k USD if the Mamiya version is $5500 USD....hell lets make it $15k - call it ultra-megha phocus lead free version - we all know someone will buy it, probably a lot of people!
[a href=\"index.php?act=findpost&pid=145025\"][{POST_SNAPBACK}][/a]

This emotional vendetta must have some sort of root cause. Care to share? Maybe we can help.