SmartVision Computer Vision Tech Comes to Image Transcoding

The use of images on the web is absolutely exploding Well... it’s actually the size of the images that is going up the most. This is a chart from httparchive showing the average number of image requests and the total image bytes over the last three and a half years:

What you can see is that the number of images has gone up slightly, but it’s really the total image size, and the resulting data transfer, that has grown dramatically.

And why not? Big, beautiful images allow online shoppers the ability to see enough intricate detail to convert and make a purchase. But big images have a downside – they can kill performance, in particular on mobile devices. That’s why Instart Logic has been focusing on letting our customers deliver great images with killer performance.

Enter SmartVision technology

Earlier this year we introduced a world's first for the web performance world – combining cutting edge computer vision technology we call SmartVision with our True Fidelity Image Streaming feature. SmartVision allows our service to gain a unique understanding of each individual image that passes through it, and it lets our Image Streaming feature make smarter decisions about how much initial image information to send down for a first fast paint of the screen in a web browser.

Taking SmartVision a step further

Today we are announcing that we have taken SmartVision technology even further and applied it to the art of properly transcoding images. In the past, outputting the right image for web and mobile application delivery has traditionally been a lot of hard work. Designers and others involved in the image publishing process have to make guesses and use trial and error to determine the right output formats, compression levels, and quality levels for different types of images. This results in outputting images at too high a quality level than actually required, and ends up negatively impacting performance as users are forced to download much more image data than necessary.

Now, by leveraging SmartVision, the Instart Logic service can automatically determine the best image compression, format, and quality levels for each individual image during the delivery process by understanding their unique characteristics and visual content. By reducing the size of the image file, the download time is dramatically reduced and images are displayed to end users radically faster than before. This breakthrough is particularly important for delivery of applications to wireless devices, which often receive images over slower, congested or unpredictable wireless network connections.

If you're interested in the technical details, our Image Scientist Parvez Ahammad will be presenting a new paper at the ACM Multimedia Conference today. Check out his own blog post, and the paper he will be speaking about.