Mar
25

Adobe’s Content-Aware Fill wows, scares me (part 01)

As far as I can tell from this video, one of the primary selling-points of Adobe’s upcoming CS5 suite will apparently be ‘sorcery’.

I can say with relative certainty than on any given work week, I use at least one part of the Adobe Creative Suite every day. I keep my ear to the ground for new info, and while I do remember hearing bits and pieces about Adobe’s new content-aware tools towards the end of last year, I never got around to looking into the details.

So as I watched this Content-Aware Fill Sneak Peek video posted by Bryan O’Neil Hughes over at Adobe I found myself shocked, slack-jawed and salivating at the immense power and potential of Photoshop’s new feature-set. (I also found myself a little scared and concerned, but we’ll save that discussion for another day and another post).

Now, while I was able to suppress my natural urge to grab the ‘ol pitchfork and run out into the night decrying witchcraft, I definitely felt an immediate need to better understand the science behind the incredible image conjuring I saw.

The Sorcerer's Apprentice

Based on research done primarily with Princeton, this appears to be the practical development of PatchMatch, a “Randomized Correspondence Algorithm for Structural Image Editing”. The paper (and associated materials) were submitted to SIGGRAPH last year, and are available at the link above.

From a cursory glance, what I’ve been able to gather is that these folks have implemented a new algorithm that uses reference points, both user-defined and automatically generated, to intelligently and quickly gather a ‘best guess’ of missing/modified content in an image and use these guesses to generate and apply a texture based on existing patterns and data within the surrounding image.

The improved algorithm’s guessing is seemingly paired with enhancements to the patching system to generate semi-random, semi-new content from what it gathers within the existing image.

The big deal here is not so much that this is possible, but that this algorithm is smart about the way it analyzes things, and so it can process its results at a speed to make it available commercially viable in a real-time environment.

In very much the same way, for Adobe’s end-user, it’s the speed that’s crucial here. As is mentioned in the video, what this tool is doing was not impossible before, it just took a lot of patience, a lot of time, and typically would result in very worn-out looking ‘cmd‘ & ‘z‘ keys.

In purely practical terms, cutting out big chunks of this time-consuming type of tedium is an obvious budgetary win – not to mention having a creative staff who looks at least slightly-less bleary eyed. But perhaps more importantly, I think this will mark a significant step towards allowing imagination and creativity to flow seamlessly and fluidly into a form of visual expression.

And that, I feel, will ultimately yield better work, as it will offer up more time and resources to the ‘creative‘ parts of the creating, and less to the preparation and the (inevitable) post-creativity cleanup.

I’ll also go ahead and make a safe prediction here that we’ll see a specific form of this tech in a significant way within CS5′s motion graphics package. This type of algorithm seems like it could drive some major motion tracking and compositing breakthroughs, and I guarantee you I will be keeping a keener eye on Adobe’s Labs in the days to come.

But…. (big but)… all that excitement aside, this tech also leaves me a little frightened – at this tool’s effect on the definition of ‘image’, at its ease of use, and its potential for abuse. This will almost certainly contribute to the mindset that we must question the validity of everything we see. We’ll dive more into depth on that next time, in part two of this post. Look for it soon!

In the meantime,  pick your jaw up off the floor and tell us your thoughts!

Reblog this post [with Zemanta]

Popularity: 9%

Leave a comment

Like?

MindoMondo on Facebook

Comment-O