Perspective correction non-destructively

[Jeremy Gilbert]

I just posted an extended rant on my blog about the need to do nondestructive, fast perspective and lens correction on images.

Can we lobby Adobe to make this happen?

http://www.springamp.com/2''9'325/fast-non-destructive-lens-and-perspective-correction-please

In summary:

1) Build a perspective correction tool that lets you manually drag across lines that should be either vertical or horizontal, computing the correct rotation, horizontal and vertical correction factors. Check out AutoPano Pro — they seem to have no problem doing this
2) Buy out Tim Niemann, the PT Lens guy or license his stuff and build the lens distortion correction right into Lightroom. No your clever Lens Correction plugin isn’t enough; distortion varies by focal length for many lens!
3) Let me apply these three filters right inside of Lightroom and/or Aperture without creating a intermediate PS file
Lens correction
USM
Perspective correction
4) Provide more fine-grained control over shadow/highlight boosting right inside of Lightroom

 

[Brad Snyder]

Jeremy, often requested since day one. The feature request link is on the NavBar at the top of the forum, if you'd like to add your voice. Be sure to explain and document why it's in the development team's interest to prioritize your request. I'm sure you already know that the entire development process is a series of tradeoffs between different user camps espousing different 'absolutely must-have' features.

Not sure what to tell you about Aperture tho'. Surely there's a better venue to discuss that?

 

[Jeremy Gilbert]

Benefits to Dev Team

Brad --

I'll let the Adobe team know, but I'd also like to hear what people on the forum say.

I've logged some serious time (yes, i've put in my "1',''' hours") as a product manager, and J2EE software architect, so I know the challenges of pushing a feature into a busy dev team. But I think in this case, its important to recognize features that add "transverse capability" vs. merely "enhancements." We also have to recognize that this product space is fast-moving, and the product will need to evolve substantially in the coming years to survive what surely will be waves of commodification of picture editing and sorting capabilities. (That last part is from working as a management consultant.)

I think if Adobe wants to be confident they will continue to have a leading product in the future, they need to rethink the paradigm used for image adjustments entirely. Rather than argue if its perspective correction next or liquify, or whatever, they need to add some transverse capabilities to the product so that all of these types of features can proliferate in a non-destructive workflow.

In this case, the LR team needs to invest in a way to get modularization into the core rendering engine of LR itself. There is always going to be a better way to transform pixels for lens correction, noise reduction, perspective correction, etc. But when you have a system that can take any INPUT image, a set of parameters, and can produce an OUTPUT, you've opened up your platform to some pretty rich capabilities.

Right now, LR appears very monolithic: its non-destructive adjustments are deeply integrated into each other such that adding one more feature requires adjusting a lot of code. All calculations are probably made in float, and we know its using ProRGB. When one adjustment lightens part of an image, an other may countervene by darkening it, so there is a lot of stacked parameterized adjustments going on. Each slider has to render fast, both immediately on-screen in low-res, but then again in high-res. Worse, any capability added here has to be rolled out across all ACR-based applications.

I think they should invest some serious time to a plug-in architecture that will allow them to deliver a lot of different features at once without overloading their program. Each plug in can describe which stage in the render pipeline it needs to fit, and offer a quality/speed tradeoff. Hell, maybe it can describe what it does in some XML formula which is then inserted in the right place in the render pipeline, and compiled to gain efficiency with the other adjustments stacked on top of it. Maybe you cache intermediate ProRGB floating point world output from the early raw decode and lengthy transforms somewhere on disk so the other parts of the UI will run fast when you adjust exposures, etc. Something VERY clever can be done to enable this functionality and Adobe should lead in this area.

Benefits to the dev team:

1) Faster responsiveness to future product requests
2) The ability to engage more developers to productively improve the product at one time; combatting the mythical man month problem
3) The ability to segregate feature production from performance improvement logic within the development team
4) Helping gain an unassailable lead in terms of hard feature capabilities in the future
5) Strengthening the entire Adobe product ecosystem with more robust leverage of 3rd party talent and economic motive

Rumor has it that on Google's campus, Serguy and Brin push product teams to "do more than the expected or conventional". I think in this case, LR needs to do the same. We need some serious "Google Gears" thinking here, and LR will be stronger for it.

Jeremy

 

[Kiwigeoff]

Brad --

I'll let the Adobe team know, but I'd also like to hear what people on the forum say.

I've logged some serious time (yes, i've put in my "1',''' hours") as a product manager, and J2EE software architect, so I know the challenges of pushing a feature into a busy dev team. But I think in this case, its important to recognize features that add "transverse capability" vs. merely "enhancements." We also have to recognize that this product space is fast-moving, and the product will need to evolve substantially in the coming years to survive what surely will be waves of commodification of picture editing and sorting capabilities. (That last part is from working as a management consultant.)

I think if Adobe wants to be confident they will continue to have a leading product in the future, they need to rethink the paradigm used for image adjustments entirely. Rather than argue if its perspective correction next or liquify, or whatever, they need to add some transverse capabilities to the product so that all of these types of features can proliferate in a non-destructive workflow.

In this case, the LR team needs to invest in a way to get modularization into the core rendering engine of LR itself. There is always going to be a better way to transform pixels for lens correction, noise reduction, perspective correction, etc. But when you have a system that can take any INPUT image, a set of parameters, and can produce an OUTPUT, you've opened up your platform to some pretty rich capabilities.

Right now, LR appears very monolithic: its non-destructive adjustments are deeply integrated into each other such that adding one more feature requires adjusting a lot of code. All calculations are probably made in float, and we know its using ProRGB. When one adjustment lightens part of an image, an other may countervene by darkening it, so there is a lot of stacked parameterized adjustments going on. Each slider has to render fast, both immediately on-screen in low-res, but then again in high-res. Worse, any capability added here has to be rolled out across all ACR-based applications.

I think they should invest some serious time to a plug-in architecture that will allow them to deliver a lot of different features at once without overloading their program. Each plug in can describe which stage in the render pipeline it needs to fit, and offer a quality/speed tradeoff. Hell, maybe it can describe what it does in some XML formula which is then inserted in the right place in the render pipeline, and compiled to gain efficiency with the other adjustments stacked on top of it. Maybe you cache intermediate ProRGB floating point world output from the early raw decode and lengthy transforms somewhere on disk so the other parts of the UI will run fast when you adjust exposures, etc. Something VERY clever can be done to enable this functionality and Adobe should lead in this area.

Benefits to the dev team:

1) Faster responsiveness to future product requests
2) The ability to engage more developers to productively improve the product at one time; combatting the mythical man month problem
3) The ability to segregate feature production from performance improvement logic within the development team
4) Helping gain an unassailable lead in terms of hard feature capabilities in the future
5) Strengthening the entire Adobe product ecosystem with more robust leverage of 3rd party talent and economic motive

Rumor has it that on Google's campus, Serguy and Brin push product teams to "do more than the expected or conventional". I think in this case, LR needs to do the same. We need some serious "Google Gears" thinking here, and LR will be stronger for it.

Jeremy

Some interesting thoughts here Jeremy. My experience shows that what we need and what we want/would like are two very different concepts. Needs are actually very few.
Basing thinking on rumour is fraught with danger as well and I'm surprised that you have knowledge of how the LR team works - I have only a little!!
May I suggest you refine your wording into a clear and concise explanation of what you would like to see and post a feature request as Brad suggested.
Look forward to some more of your input!!

 

[dj_paige]

I've logged some serious time (yes, i've put in my "1',''' hours") as a product manager, and J2EE software architect, so I know the challenges of pushing a feature into a busy dev team. But I think in this case, its important to recognize features that add "transverse capability" vs. merely "enhancements." We also have to recognize that this product space is fast-moving, and the product will need to evolve substantially in the coming years to survive what surely will be waves of commodification of picture editing and sorting capabilities. (That last part is from working as a management consultant.)

I read this far and I realized I have no idea what this means. I don't even know who "we" is that you refer to in the last sentence before the parenthesis. I have never seen the word "commodification" before, and although I have seen the words "transverse" and "capability" before, I have no idea what they mean together.

So although it may be "important to recognize" these things, I don't recognize these things. It might also be true that we "have to recognize" that the product space is fast-moving, I don't recognize this.

And while I would like to support your original position regarding lens perspective correction, I simply don't see the urgency that your appear to see, nor do I recognize the things you say we have to recognize.

 

[Jeremy Gilbert]

Feature Request:

(1) Put in some sliders in the develop module that let me perspective correct and lens correct. Add USM if you can.
(2) Do this in a way that 3rd parties can add more features like this

Rationale:

- Happier users
- Faster development time for your next product revs
- A more advanced product that will withstand new entrants in the market because it can be enhanced more quickly

Consequences for not doing (1) and (2):

- There are lots competitors in this area of the market and Adobe will get trumped by Picasa, Iphoto and Microsoft in the mass market if it can't be a few steps ahead

 

[Denis Pagé]

Welcome to the Lightroom Forums Jeremy.

Reading your posts and replies you got attentively, it appears to me that you think that you are actually writing to the Adobe Staff. Even if the Lightroom gurus here are Adobe Photoshop Lightroom professionnals, this site is privately owned and not connected with Adobe nor sponsored by Adobe. Altough, some Adobe representatives may login from time to time and we gave them a badge as "Ar".

When Geoff Walker asked you to reword and post a feature request, he didn't mean to write back here with more details but to use the "Log Bug/Feature Request" link we provide on the blue bar top of the forums. THIS ONE is going to send your requests to the real Adobe staff. They may not give you a feedback but they will certainly read it! The more users asking for a given feature, the better are the priorities.

That said, feel free to tell us what you wrote them but I think we already got a clear idea. We all want perspective adjustments at some point in the future...

 

[pacorocha]

We all want perspective adjustments at some point in the future...

Sorry Denis. "perspective adjustments at some point in the NEXT future...:mrgreen::mrgreen::mrgreen: