Wednesday, February 10, 2010

Analyzing art by the brush stroke

Scan in an image of a drawing or painting. Determine the colours in use. Enhance the image. Analyze the path, density, length and style of brush or pen strokes and facial pattern recognition. Add the "textons" or patterns to a lexicograph catalog and scan for similarities. Search for identifying features that "watermark" the painting or sketch, like finger strokes or brush bristles that make it similar to other catalogued art. Sceptical yet?

Add further features to a scanner to detect the "depth" or pressure used in a brush or pen stroke. Analyze its chemical composition with a light non-intrusive spray mist, to determine the composition of chemicals paints used, and highlight materials or chemicals consistent with a certain time period.

Reverse or present a negative grayscale view of the image, to determine what "isn't" there, such as cracks or craquelure and features that are not part of the painting. Measure these features. Treat them like fingerprints that grow as the art ages.

Take it further and analyze the signatures of the scanned image. Wavelets. Pixel composition. Orientation and scale of the strokes.

Then use a 3D oil paint printer to spit the image back out, perhaps removing the age to reinvent the image.

The program determines whether a particular drawing is consistent with an artist's style. Until now, Rockmore has only tested his program on Bruegel drawings, but he says there is no reason it could not be used for other artists.

The original paper here.

How long before we go from just rendering an actor digitally on a computer screen to building an application that generates random Rembrandts? Or how about the Harry Potter Sequel Generator? What about automatically generating books "in the style of" an author, with programs that do more than just combining common phrases the author uses, but actually analyze patterns in the writing to formulate new plot lines and characters?

Give it 5 years or so... let's check back in 2015 once cloud computing, oData, GPU server farm hardware, 4d scanners and 3d printers mature.

Beej explains it best, as always.

Let me start off by saying, “Don’t do any of this yourself.” This is one of those topics where there are countless libraries already written for you with well-designed and well-tested code to do everything you want and more. It’s presented here because it fits criterion #1 for why you’d re-invent the wheel: you’re doing it because you want to learn more about it.

It's amazing that all of the techniques, software and hardware are already available to scan in and reproduce a near-perfect copy of a Great Masters artwork.

What's missing?

What was the artist thinking when the picture was created?

Truly, an artist's expression within their art is impossible to perfectly reproduce by another artist. Hence the ability to detect forgeries. The talent and expression of an artist is the true component that cannot be reproduced.

But, like compressed sound files, will people really care that the reproduction is missing things outside of their field of vision?

In my opinion, the original is still better than a reproduction when it comes to art.

What would you rather do, listen to a musician in your car or in an arena? Download art on your computer or visit a gallery? Looks at someone else's photos of a trip or actually visit the place yourself? Watch Avatar in IMAX 3D on a 72 foot screen or just Blu-ray on a 32" TV?

Drive a Lamborghini or a Tractorri?

That last one was a trick question. I'd take the Tractorri. Now that is art. Quite possibly the most disturbing art I have ever seen.

No comments: