Please subscribe to my feed.

You can subscribe by entering your email address below, and you will never miss any good posts by our panel of authors. Don't worry, you can unsubscribe ANYTIME.

Should you want to join Kongtechnology.com, please read here.

5 researchers from China and Singapore have given us a glimpse into what is possible in the future of image editing when they stunned the world by releasing Photosketch, a software which has almost magical abilities in its simplicity, yet is based on probably some very complicated stuff.

Photosketch is an elegant concept: you sketch a scene of, say, a dog chasing after a robber using your mouse. Then the next step is to label your sketch, eg ‘Dog’ and ‘Robber’. With this information input into Photosketch, the output will be a photo of what you drew, except all you did was do a crude sketch while the output is a beautiful picture. See the demo video here.

Helicopter

For professionals, this is the art of image editing, or also known as photoshopping. However instead of having to hold a degree in design, all you need is this tool and a little imagination; the rest is automatic.

Photosketch works by pulling images from the internet (Google Images, Bing Image Search, etc). From this vast library of photos, it determines what you need. After it has gotten the sample pictures, it is then composited together. It is clever enough to stitch individual elements into a single coherent figure, even adding shadows to make it more realistic. It also uses fancy statistical algorithms to filter out incorrect or undesirable images leaving only the best images to work with.

Shark

The appeal lies in its simplicity. No doubt big companies like Google or Microsoft will be quick to investigate and perhaps invest into the technology, by which time it will be available to the public. For now, Photosketch remains a research in progress, with more work to follow.

One thing which remains to be seen is how far the program is from final release. Some people at Slashdot have commented that the tool is not as ‘smart’ as it’s thought, requiring tweaks to its configuration files to set the number of ‘objects’ and also relying on a library of  potential matched images into a directory for the tool to access and analyze; this is a far cry from the ‘able to search the whole Internet’ claim. Nonetheless, it works, and with the floodgates open, more development is sure to follow.