Extrapolate letters from blurred writing (e.g. license plate)?

My friend has a blurred image of a thief's license plate. Is it possible to run an algorithm on these pixels to determine the most likely characters that the p开发者_StackOverflowixels represent?

(The fact that it's a license plate is irrelevant, the solution should work by principle on any photographed text that is difficult to decipher.)

Please help me find my dog!

Update: My friend sent my two still images, they are very poor. He doesn't have any shots of the license plates unfortunately.

Image 1: An image of the van. I'm trying to decipher the text above the word 'rentals'.

Image 2: This is the same van, with presumably the same text. It's in the upper right-hand corner of the image

It depends on why/how it's blurred. There are a number of things you could try though: one would be a simple sharpening with an unsharp mask. Another I've found surprisingly effective at times is to simply invert the colors in a photo -- sometimes things that are really hard to read normally just pop right out when inverted.

For a one-time task like this, however, you probably want to use existing tools (e.g., Photoshop or The Gimp), rather than writing new code. It'll take a long time and a lot of effort to match what they already have just waiting to be used.

Play with Photoshop. Try different sharpening filters, in different strengths and different orders. Also play with posterization. Revert to the original image frequently. Look for what works. Use your eyes. If you can't see the answer (after applying filters), OCR probably won't either.

Motion blur can be removed, because all of the information is still in the photograph. But in this case, I'm not sure any form of image processing is going to help.

I apologize if you already tried this, but have you looked through the "rentals" section of the phone book to see if you can find a company with a similar logo? I assume this is a van from a rental equipment or "rent-to-own" business, not a car rental agency.

I'd recommend a sharpening, followed by Sobel filter to find edges, then perform your OCR on it.

Refs:

http://en.wikipedia.org/wiki/Sobel_operator

http://www.bythom.com/sharpening.htm

Theoretically it is possible under ideal conditions. But it requires that you know the transform from the original to the blurred image.

Image compression, non-linearities in the camera, limited resolution and noise may get in the way. If you're lucky a standard photoshop sharpening filter will do.