Microsoft Research creates HDR hallucination (without drugs) | The Coffeehouse

longzheng

A few weeks ago the Microsoft Research TechFest was on, and everyone feasted their eyes on some of the amazing technologies MSR had to show off. But there was one that didn't get mentioned, or at least I don't it did, otherwise I'd look rather foolish.

Anyhow, the technology is called “High Dynamic Range Image Hallucination” and is as good as hallucinations go. In a nutshell, if you like nuts, they can recreate detail lost from low-dynamic range photographs to create high-dynamic range photographs.

The result? Over-exposure and under-exposure is simply a few brushstrokes to fix. This should appeal to photographs all the way from amatuer to professional, since everyone makes mistakes.

Here's a few screencaps.

Texture sample (left). Original image (middle). Original image with ‘Hallucination’ processing (right).

High Dynamic Range Image Hallucination
Original image (left). Original image under-exposed (middle). Original image under-exposed with ‘Hallucination’ processing (right).

Check out my blog post on it with a YouTube video of their presentation. Smiley

RoyalSchrubber
One. How many time travellers does it take to change a lightbulb?

wow, thanks for this TechFest report, longzheng.

MS Research once again proved they totally rock Smiley

Raghavendra_Mudugal
This is how it feels when you do not upgrade to the latest technology...

WOW, thanks for this coooooooooooool stuff. [A]

Lloyd_Humph
If Blackberrys are addictive cellphones, Channel9 is the ultimate addictive website.

MSR really do rock! now then...

ZippyV
Fired Up

It's sad that we can't download the program and play with it, just like Photosynth. Lot's of interesting things but we can't play with it.

Why do they even release any papers if we can't play with it?

longzheng

ZippyV wrote:

It's sad that we can't download the program and play with it, just like Photosynth. Lot's of interesting things but we can't play with it.

Why do they even release any papers if we can't play with it?

Because this is Microsoft Research, not Microsoft.

thumbtacks2

longzheng wrote:

ZippyV wrote: It's sad that we can't download the program and play with it, just like Photosynth. Lot's of interesting things but we can't play with it.

Why do they even release any papers if we can't play with it?
Because this is Microsoft Research, not Microsoft.

Aw, come on.

Sad

I do wish more researchers would post here, though. Why don't they?

By the way, although Photosynth is interesting, but it doesn't take a lot of work to write a program that let's you flip and display photographs in 3-D (I was able to do that last night in fact). I wonder how much work it would take to come up with some useful searching algorithms...

Harlequin
http://twitter.com/TrueHarlequin

thumbtacks2 wrote:
By the way, although Photosynth is interesting, but it doesn't take a lot of work to write a program that let's you flip and display photographs in 3-D (I was able to do that last night in fact).

Photosynth flips photographs in 3D......did you even look at the thing? Look at the videos on it? Perplexed

I thought version one of thumbtacks was pathetic enough. Was hoping a 2.0 would be some improvement.

ZippyV
Fired Up

thumbtacks2 wrote:
By the way, although Photosynth is interesting, but it doesn't take a lot of work to write a program that let's you flip and display photographs in 3-D.

This is not the most interesting part of Photosynth. It's the application that puts everything in a 3D environment based on those pictures. I'd like to take some pictures myself from my town and see how the results from that. I'm not interested in the places they've shown on their site.

thumbtacks2

Harlequin wrote:

thumbtacks2 wrote: By the way, although Photosynth is interesting, but it doesn't take a lot of work to write a program that let's you flip and display photographs in 3-D (I was able to do that last night in fact).
Photosynth flips photographs in 3D......did you even look at the thing? Look at the videos on it?

I thought version one of thumbtacks was pathetic enough. Was hoping a 2.0 would be some improvement.

?

Yes, I watched the video, thanks. And looked at lots of screenshots. No, I haven't installed it, though. I realize there are significant algorithms behind it, and I do realize its intent. I thought it was an interesting idea. Is it so original that it cannot be done by anyone else? No, I don't think so.

But like Paolo said, I need to "focus" on one project at a time.

thumbtacks2

Come to think of it...maybe I'll stop posting here, too, for a while.

Like a real long time.

It just gets tiring how you guys sit and praise the Research department like everything they do is groundbreaking. It really isn't sometimes. But I think it would be better if I continue to spend most of my energy proving that point in terms of code and not wasting my time here.

C'ya.

Harlequin
http://twitter.com/TrueHarlequin

I think all research departments are groundbreaking, in every industry on the planet. That's what makes them research departments Cool

Bas
It finds lightbulbs.

thumbtacks2 wrote:
Come to think of it...maybe I'll stop posting here, too, for a while.

Like a real long time.

It just gets tiring how you guys sit and praise the Research department like everything they do is groundbreaking. It really isn't sometimes. But I think it would be better if I continue to spend most of my energy proving that point in terms of code and not wasting my time here.

C'ya.

Until tomorrow!

thumbtacks2

Bas wrote:

thumbtacks2 wrote: Come to think of it...maybe I'll stop posting here, too, for a while.

Like a real long time.

It just gets tiring how you guys sit and praise the Research department like everything they do is groundbreaking. It really isn't sometimes. But I think it would be better if I continue to spend most of my energy proving that point in terms of code and not wasting my time here.

C'ya.
Until tomorrow!

No, I'm serious. I have lots of great ideas I'm in the process of implementing. It's just a shame that when MS researchers or other researchers from other companies do post on this site you guys manage to either ignore them or let the conversation pretty much tank. Heck, it looks like you even got Miguel de Icaza to post here and the reaction was really flat. What's up with that? I remember a search engine guy coming here a ways back who built a search engine out of .NET. Not that I found that too exciting being .NET and all that, but the fact that he built one was interesting. That takes significant skill. It's just really weird to watch this happen day in and day out, and it probably explains why more research types don't show up here and probably never will.

Sourcecode
Whatever it is, I didn't do it.

thumbtacks2 wrote:

Bas wrote:

thumbtacks2 wrote: Come to think of it...maybe I'll stop posting here, too, for a while.

Like a real long time.

It just gets tiring how you guys sit and praise the Research department like everything they do is groundbreaking. It really isn't sometimes. But I think it would be better if I continue to spend most of my energy proving that point in terms of code and not wasting my time here.

C'ya.
Until tomorrow!

No, I'm serious. I have lots of great ideas I'm in the process of implementing. It's just a shame that when MS researchers or other researchers from other companies do post on this site you guys manage to either ignore them or let the conversation pretty much tank. Heck, it looks like you even got Miguel de Icaza to post here and the reaction was really flat. What's up with that? I remember a search engine guy coming here a ways back who built a search engine out of .NET. Not that I found that too exciting being .NET and all that, but the fact that he built one was interesting. That takes significant skill. It's just really weird to watch this happen day in and day out, and it probably explains why more research types don't show up here and probably never will.

Perhaps but with condescending comments like..

thumbtacks2 wrote:

By the way, although Photosynth is interesting, but it doesn't take a lot of work to write a program that let's you flip and display photographs in 3-D (I was able to do that last night in fact).

Don’t expect to many ears and mouths moving in a favorable direction toward you.

What is probably an enormous amount of research and work, over the course of a large time span being reduced into a single night of work by yourself; doesn’t provide merit toward the validity of what your saying.

Bas
It finds lightbulbs.

thumbtacks2 wrote:

Heck, it looks like you even got Miguel de Icaza to post here and the reaction was really flat. What's up with that?

I remember Miguel de Icaza coming here followed by many people welcoming him here and asking him questions about Mono. I also remember people asking him to post here more often. I'm not sure how that qualifies as flat.

thumbtacks2

Sourcecode wrote:

thumbtacks2 wrote:
By the way, although Photosynth is interesting, but it doesn't take a lot of work to write a program that let's you flip and display photographs in 3-D (I was able to do that last night in fact).
Don’t expect to many ears and mouths moving in a favorable direction toward you.

What is probably an enormous amount of research and work, over the course of a large time span being reduced into a single night of work by yourself; doesn’t provide merit toward the validity of what your saying.

I was not trying to be condescending. My point was that over a few nights, I found it somewhat easy to put together an application that reads in bitmaps off a disk, pastes them as textures onto a series of quads, and then spread them out onscreen like photos on a table. From there I was able to add some basic keyboard code to allow you fly over the photos and zoom in on them. It wouldn't take much more work to add mouse functionality, allow those photos to be flipped over and even sort them into a photo-album type structure. The real work would be build a nice sorting algorithm, something that can sort through the photos in a useful manner, etc.

And yes, I do understand Photosynth does a lot more than that. I'm not knocking Photosynth; but what I am saying is that somebody with enough time and devotion can come up with equally impressive software without a huge budget...although a huge budget would definitely help I'm sure.

dentaku

That's looks very useful. I could see myself using that for outdoor photographs.

JasonOlson

thumbtacks2 wrote:

And yes, I do understand Photosynth does a lot more than that. I'm not knocking Photosynth; but what I am saying is that somebody with enough time and devotion can come up with equally impressive software without a huge budget...although a huge budget would definitely help I'm sure.

Did I miss somewhere that says Photosynth took a huge budget? IIRC, it was basically a Senior Thesis by a student at University of Washington that he was doing at an internship with Microsoft Research (or somethign like that (I could be WAYYYYYY off and thinking of something else)).

Of course, that kind of shows your point. But, why _hasn't_ anyone else productized it? Why haven't we seen it out there before. Isn't the whole thing about "revolutionary" that something is developed that has never been seen before, NOT how much effort was put behind it?

Anyways, I may just be mis-understanding what you said. If so, my apologies.

thumbtacks2

JasonOlson wrote:
Did I miss somewhere that says Photosynth took a huge budget? IIRC, it was basically a Senior Thesis by a student at University of Washington that he was doing at an internship with Microsoft Research (or somethign like that (I could be WAYYYYYY off and thinking of something else)).

Okay...but usually some funding is involved in the research area...or so I thought. Also, was he hired for his efforts?

JasonOlson wrote:
Of course, that kind of shows your point. But, why _hasn't_ anyone else productized it? Why haven't we seen it out there before. Isn't the whole thing about "revolutionary" that something is developed that has never been seen before, NOT how much effort was put behind it?.

Don't get me wrong, I think it has a lot of uses. Those are what make the best products...ideas that are "revolutionary" yet useful.

Richard.Hein
... my guitar gently weeps ...

thumbtacks2 wrote:

It wouldn't take much more work to add mouse functionality, allow those photos to be flipped over and even sort them into a photo-album type structure. The real work would be build a nice sorting algorithm, something that can sort through the photos in a useful manner, etc.

And yes, I do understand Photosynth does a lot more than that.

No, you clearly do not understand Photosynth. Photosynth does have a "nice sorting algorithm". The way that points are analyzed and sorted and mapped out in space to allow the system to tell that two photos are related, and know how to lay them out in 3D in a plane - YOU have no idea what you are talking about. A lot of people DREAM about ideas like this, I know I have, but to actually code it. Ha. Man, you just have to smarten up and watch the Photosynth technical video interviews, because you clearly don't know what Photosynth's capabilities are.

PaoloM
Hypermediocrity

thumbtacks2 wrote:

Bas wrote:

thumbtacks2 wrote: Come to think of it...maybe I'll stop posting here, too, for a while.

Like a real long time.

It just gets tiring how you guys sit and praise the Research department like everything they do is groundbreaking. It really isn't sometimes. But I think it would be better if I continue to spend most of my energy proving that point in terms of code and not wasting my time here.

C'ya.
Until tomorrow!
No, I'm serious. I have lots of great ideas I'm in the process of implementing. It's just a shame that when MS researchers or other researchers from other companies do post on this site you guys manage to either ignore them or let the conversation pretty much tank. Heck, it looks like you even got Miguel de Icaza to post here and the reaction was really flat. What's up with that? I remember a search engine guy coming here a ways back who built a search engine out of .NET. Not that I found that too exciting being .NET and all that, but the fact that he built one was interesting. That takes significant skill. It's just really weird to watch this happen day in and day out, and it probably explains why more research types don't show up here and probably never will.

In the meanwhile, we'll twiddle our thumbs waiting for your vapourware.

Remember, to be able to convincingly criticize something, you should have been able to come out with something of equal or above importance. Anything else is just idle chit chat.

Rory
Free Tibet While Supplies Last

thumbtacks2 wrote:
It just gets tiring how you guys sit and praise the Research department like everything they do is groundbreaking. It really isn't sometimes. But I think it would be better if I continue to spend most of my energy proving that point in terms of code and not wasting my time here.

1. There's nothing tiring about prasing MSR - it's certainly better than the alternative.

2. An idea doesn't need to be complex to be groundbreaking. Saying that you could copy an MSR project isn't the same as coming up with the project. If an idea's great but only takes a few hundred lines to implement, it's still a great idea.

3. If you want to prove a point, then do something "groundbreaking" on your own - not just whipping up a project based on something done by MSR. Your own idea.

And I'm not being sarcastic or biting here. Quite serious.

Given that this site is a community for devs interested in Microsoft, it makes sense for us to get excited about Microsoft projects. Doesn't mean other projects aren't exciting. It's not a contest.

thumbtacks2

PaoloM wrote:
In the meanwhile, we'll twiddle our thumbs waiting for your vapourware.

Remember, to be able to convincingly criticize something, you should have been able to come out with something of equal or above importance. Anything else is just idle chit chat.

Vapourware? Last night, as I already stated, I wrote up the 3-D photo display thing that allows you to zoom in and out and fly around above a bunch of bitmaps. Not difficult really, and certainly not a finished product. But the game engine is moving along just fine, thanks. But I believe it was you who stated that I should "focus" and another person said I should only post "finished products".

PaoloM
Hypermediocrity

thumbtacks2 wrote:

PaoloM wrote: In the meanwhile, we'll twiddle our thumbs waiting for your vapourware.

Remember, to be able to convincingly criticize something, you should have been able to come out with something of equal or above importance. Anything else is just idle chit chat.
Vapourware? Last night, as I already stated, I wrote up the 3-D photo display thing that allows you to zoom in and out and fly around above a bunch of bitmaps. Not difficult really, and certainly not a finished product. But the game engine is moving along just fine, thanks. But I believe it was you who stated that I should "focus" and another person said I should only post "finished products".

Last night I built a perpetual motion engine. Not difficult really, and certainly not a finished product.