Getting closer to the 'enhance' button
Photoshop may soon be able to correct focus and motion blurring
This has become such a common plot device that it barely requires explaining anymore. The investigator/CSI/forensic technician obtains a surveillance video clip that includes the barest remnant of the elusive criminal’s identity. Pointed out the entry door of a stop-and-rob, the surveillance camera picks up the image of the fleeing escape car, reflected in the chrome hubcap of a pickup truck across the street. “Enhance that section for me,” our hero asks. A few key-presses later, the onscreen blobs resolve into a perfect image of the bad guy’s license plate, or sometimes even a portrait of the evildoer himself.
At a recent trade show hosted by Adobe — makers of the widely-used Photoshop software — a software engineer demonstrated an in-the-works feature that will probably show up in the next edition of the program, or maybe the one after that. It’s not quite as good as the TV detective’s “enhance” button, but it represents a remarkable accomplishment in the correction of blurring, due either to poor focus or to motion.
Historically, blurring has been the one photo problem you can’t correct in the darkroom. Too much or too little light, within limits, can be fixed by “pushing” the conventional film in the developer, or by altering the brightness and contrast settings in a digital image. If you didn’t focus the camera correctly, or you lacked a sufficiently steady hand during a long exposure, you got a blurred picture, and that was that—there’s no way to fix it.
In the demonstration that was recorded on a camcorder by conference participant Peter Elst at the 2011 Adobe MAX awards in Los Angeles last week, an engineer demonstrated three applications of the new feature. In the first example, a photograph of the interior of a shopping mall had been blurred by camera motion during the exposure. The software first analyzed the image to determine the motion pattern, then corrected it to bring the photo into sharp focus. A second example took a cell phone camera shot of a bulletin board poster.
Like a lot of cell phone pictures, it’s too close to focus sharply and most of the text is unreadable. Click, click, and the text snaps into focus. The third image was of Adobe’s CEO, speaking at a conference. The photo is low-resolution and fuzzy. A few clicks eliminated the distortion and sharpened it. If you view the demonstration clip linked above, expand the window to full screen. The effects are hard to see in the typical YouTube-size window.
Enhancing a photo TV-style usually isn’t possible because the information simply isn’t there to extract. When we look at a photograph, we’re actually seeing an arrangement of dots of varying colors. With conventional film, the dots are the “grain” of the silver halide particles on the plastic substrate; in a digital image, they’re pixels. The maximum number of pixels is determined when the image is recorded. In a standard video frame, there are 640x480, or 3,072,000 pixels to work with. If the license plate, face, or purloined letter in question is covered by only a few of those pixels, “enhancing” the image is going to produce only blobs of color.
When the image is blurred as in the Adobe examples, the desired information is there, but it’s muddled with other information surrounding it. Given the right instructions, the computer can find the key elements and reconstruct the edges of the image, producing a sharp picture. This kind of magic is possible mostly because today’s computers are so much faster than those from a few years ago. With multiple-core processors and increased clock speeds, a midrange desktop computer of today is equivalent to one from ten years ago that was the size of two refrigerators and used more power than an office building.
Adobe wasn’t promising that the new blur correction feature would be in the next version of Photoshop (which should be version “CS6”), or even the one after that. Adobe doesn’t usually hold onto new features for long, though. A couple of years back they demonstrated a feature called “content-aware fill” that allowed a user to select an unwanted feature — say, a fence — in a photo and the software would not only delete it, but filled in the missing space with details from the surrounding background, making it difficult to tell there was ever anything there. That feature showed up in the next iteration of Photoshop, where it remains today.
Soon, you may be able to ask, “Enhance that section for me,” and actually get something worthwhile.
|Back to previous page|