Hopefully it will get to the point where we can cut out an image, then do:
Match Light Source
and noone would be able to tell that any manipulation was done at all!
»Match Light Source«?
Aren’t You expecting a bit much from an image-editing program because light-sources usually would have to be considered three dimensionally or was this proposed jokingly?
Actually it's not as crazy as it sounds.
Listen to the Interview with Russell Williams(Co-Photoshop Architect) at http://www.photoshopcreative.co.uk/podcasts.php
He talks about the future of photoshop and says their labs are actually working on removing light sources from images right now (with some progress already!). He says his main goal is to eventually cut out an image, remove the light source, paste it into another image, and give it the light source from that image (match light source).
Photo manipulations would never be the same if photoshop were to successfully integrate this functionality.
*i'm pretty sure its near the end of the podcast. (you don't have to listen to the whole thing, but you should, its interesting stuff)
Thanks for the link.
But does he not cite that as a "damn-the-costs", ”wishful-thinking" kind of feature?
To achieve this the program would have to be able to make educated assumptions on the 3-dimensional form of elements in the image and recognize various areas as belonging together to begin with.
And then there are the shadows that various objects cast onto one another …
So I would guess that’s a feature that may be quite a ways off yet.
Regarding interesting links:
It’s indeed fascinating, what can be programmed, and this appears to be a couple years old.
But that program seems to focus on the head alone, and I think the rest of the body (and textiles) might still pose a couple challenges.
One reason I think people would expect full three dimensional editability with the change of light-sources is to correct the possible perspectival discrepancies of the elements they combined.
This all sounds like complete gobbledegook to me. How on earth does a 2D piece of software analyse existing light sources, texure and planes- all that would ne necessary to convincingly change the light source in a flat 2D image. Even the abilty to map complex geometric planes and surfaces from a simple 2D surface is just not going to happen, for a whole variety of practical reasons. I think it may be an "ideal" of what should be possible, what would be an ideal feature. Its good to know that there are people within the Photoshop team still thinking in these idealistic terms.
Well, it may just be short of gobbledegook … or maybe not.
But a sufficiently complex 3D-analysis/modeling-program that incorporates an humongous library of statistically evaluated 3D-objects might get some results (see the youtube-link regarding automated construction of facial 3D-models from a single 2D-image).
Of course that would mean a whole different beast than what Photoshop is.
And I don’t expect that to happen anytime soon, either.
Ahh, it seems nobody wants to comment on the actual topic!
But the actual conversation going on is much more interesting.
Another interesting method to accomplish a light source match would be to use 2 "similar" images with different light sources.
For instance, I want to photoshop someone's head onto another person's body. Photoshop could simply impose the light source of the original persons head onto the one i'm photoshopping in (some sort of blending teqnique to preserve all the facial features of the newly added head).
In this instance the program wouldn't even need to create fake lighting sources and since you would already be using similar head perspectives (to make the manip believable), the shadows and light sources should come close to matching.
I know this isn't what I was originally talking about but this functionality would be extremely useful.
Also, you guys keep on saying this would require 3d functionality but I think that would only be true for adding a light source to an image. If you were just removing a light source, i feel like there could be some sort of brightness/color averaging instead of creating a 3d model, detecting the light source, and removing it from the 3d model.
The combination of these two tools would be amazing! It wouldn't be a complete match light source but it would be close.
Listen to me ADOBE!!!!!
also, someone please comment on my match quality idea (Does noone else want this?).
Ok your word "quality" - I assume means the internal texture of an image? So I guess this could be expressed as noise structure, frequency, or even jpeg atifacts.
The technology to analyse these qualities in an image amost exists already when you think about it, with the much researched and extremely powerful healing technology. Could this be stretched to try and extract an "average" texture - and apply this to another image? Yes I think it could. A similar request was made by me once.
I would settle for the functionality of the jpeg compression. As I was describing in my first post, I have to save an image with lower jpeg quality before cutting it out. The real problem with this is my “cut out” lines are not the same quality as the rest of the image(which I just compressed).
Since it may be awhile (if ever) before Adobe implements this functionality, do you know of any techniques that would allow me to lower the quality of an image while saving it with a transparent background?
Perhaps by saving it as a different file type that allows compression and transparency? Like a .png or something? Also please no recommendations for file types that when compressed or saved create the super jaggy edges (like a gif).
Sorry to turn this into a question thread but it would seem a waste to post in another section now that we’re on the subject. Arigato
The only applications for this feture I see is faking of photos taken from the internet or unlicensed use of material taken from the web.
Why would adobe want to support that? This would be a real pain for copyright holders and news/celebrity industry.
Here's a perfectly realistic, and most likely common, scenario.
I just bought two images from istockphoto. They have different quality. I wish there was a tool to match the different qualities of these two legally purchased images.
Also, whether Adobe wants to officially support modifying images taken from the internet or not, that is what the majority of people use photoshop for. You could speculate that most of these people are not actually paying for photoshop but I feel this is the starting point for many users who now pay for the product. The more people adobe can get interested in photoshop, the better it is for them. And adding tools to create more realistic manips from internet images is bound to get people interested.
Half images on microstocks like iStockPhoto etc are taken with point-and-shoot cameras and most of the other half with entry level DSLRs. 5% or fewer with pro-level DSLR or MF cameras. All off them have different quality. There can be a slight difference due to artifacts from lower level jpeg compression, but the most of the artifacts comes from low quality point-and-shoot camera sensors. Do you want Photoshop to introduce sensor artifacts to an image?
Your best bet is to download images one size larger than you need them, do your editing and then scale them down to the size you need.
Why do you want to match your work to lower quality?
It would be nice if there was an easy and cost effective way to download images one size larger than I needed them, but after years of graphic design, I can tell you this is not the case.
Also, lower quality doesn't neccesarily mean "bad" quality. In most cases, the lower quality image is acceptable for what is required.
I mean, even when talking about scaling down higher res images, there are numerous cases where the quality still won't match up. And if you are doing professional photo manipulation this method will just not cut it.
It would be nice if photoshop could do something as specific as adding sensor artifacts but that's seems pretty complicated.
I don't know exactly what sensor artifacts look like but if I had to guess I would say they probably look very similar to artifacts of the jpeg persuasion.
Actually, I would even be happy with a jpeg image compression slider "inside" the program.
I know the resident photoshop employee doesn't like hearing that a feature would be easy, but really, I don't see how this would be hard to add. I mean jpeg compression is already part of the program, all they would have to do is migrate it "inside" the program from the save dialog box.(excluding the possibility of preserving transparency).
Not taking a shot at the photoshop team. I seriously bow down before your greatness (seriously).