Two identical photos placed side by side. Both are are blown up to large size on the wall. The difference? One has been compressed by 70% using a specialized compression technique called NFO, while the other is the original. Can you tell the difference?
This was the challenge we offered participants at the recent Gestalt IT Tech Field Day. Calling it the “NFO Challenge,” we at Ocarina asked the attendees to pick the optimized photo from a series of image pairs. The participants in the event were some of the sharpest folks around–a group of independent tech blogggers from around the world–but we stumped them. In fact, the winner of the contest only guessed 50% correctly. It was that difficult.
NFO stands for “Native Format Optimization.” It’s a workflow option that Ocarina provides for image and video customers to shrink their data while retaining the native file format after optimization. Today, George Crump has a thoughtful piece, sponsored by Ocarina, that gets into detail about the technology on his Storage Switzerland blog.
Here are the basics: a JPEG comes in, a smaller JPEG goes out. Right now we’re supporting JPG and GIF, and we have beta support for PNG and h.264 video. Although our GUI offers a “volume knob” (which goes to 11 by the way), the standard settings shrink media files by 30%-50% with no perceivable quality loss.
To be clear, this is lossy compression in the technical sense. There’s no substitute for looking at the real thing. We’ve had plenty imaging experts at our customers tell us that the changes made by our algorithms really don’t impact quality. But if you need bit-for-bit lossless compression, we provide that through our standard compression offering.
NFO optimization introduces image changes that render an image that is visually identical. The most common method for validating this is to print some really big before and after images and go through them in excruciating detail to look for artifacts. You can calculate PSNR and other quantitative measures, but there’s no substitute for looking at the real thing.
The customers that have really taken off with this technology include folks not so much concerned with storage, but companies trying to save money on bandwidth. Internet photo sites are an obvious sweet-spot, but some less obvious users are large Internet retailers, whose bandwidth bills are dominated by product imagery. These guys are really excited to start building out with Ocarina. And that’s just the economic benefit of NFO. The ultimate benefit is improved end-user experience and loyalty when an image-laden web site loads faster. Now we’re adding some serious value to Internet brands.
Here’s the secret sauce: Ocarina’s software looks for ways re-encode the media in ways that align image encoding with the sensitivities of the human visual system. We’ll generally analyze the image in the pixel domain, making adjustments in de-noising, Luma/Chroma quantization, Huffman optimizations, better motion compensation, and more. To read about these optimizations, and hear more about what our team of Ph.D.s has accomplished with these algorithms, read our new white paper at OcarinaNetworks.com (http://www.ocarinanetworks.com/images/resources/nfo.pdf)
