If you are living in the internet world, like most of us in e27, you would have definitely come across this beautiful yet simple video of an application that turns a sketch into a real photo by stitching together various pictures from websites. Just draw a rough sketch of sailboat, birds, beaches and wedding gown and there you get the picture-perfect or should I say perfect picture.  Here, watch it to refresh your memory again.

Sketch2Photo: Internet Image Montage from Tao Chen on Vimeo.

The video got featured across various popular portals like Lifehacker and Gizmodo. After loads of difficulty, we were finally able to get hold of Dr.Ping Tan, One of the main man behind this intuitive app. Dr. Tan is an Assistant Professor in the Department of Electrical and Computer Engineering at National University of Singapore. In this exclusive interview with e27, he shared about some of the technology behind this application.

Imaging Technology

Dr.Tan’s primary research interests has always included image based concepts. The technology behind Sketch2Photo involves two different components: Image search and Image stitching.

Image search consists of querying the myriad search engines that include Google, Bing, Yahoo and even Flickr. After which a filtering mechanism sets in, to remove the unwanted images. This is where Dr. Tan mentioned that in this application there are high false negatives and low false positives i.e the probability of the right images being discarded is higher than the wrong images being used.

Hybrid Blending

After the image search, comes the image stitching part which he refers to as hybrid blending. He capture images of similar lighting condition and illumination properties to blend cohesively. This particular technology generates numerical measurement for the final blending result and this blending score is used to optimize combination of different images.

Application Status

With this technology, he is not planning to make it into a software application as of now and has only recently filed the patent for it and has no motivation to commercialize it. When asked about the current status of the project, he mentioned that right now he just has “messy” research code and his students are working on a live demo of the software.

I for one can’t wait to sketch this and see what turns out :P