CG in Asia
July 2, 2013, Siggraph, Anaheim, CA—In a series of small-scale presentations, Asian researchers displayed their efforts.
Yoon Chung Han, studying at UC, Santa Barbara presented Digiti Sonus, a software application that transforms a fingerprint into a collections of sounds. The selection of fingerprints is due to the relative uniqueness of fingerprint patterns among the various possible body patterns. The research goes into the intersection of media, art, and technology.
The app scans a finger to get a print, which is expanded into a 3-D image. A feature extraction process captures the minutia which can vary with the person’s pressure and angles of application. The algorithm creates a greyscale image that indicates position, height, and width of the details. The features have lines and spaces, and a set of 5 of these features is collected for analysis.
The program then traverses the image and transforms the graphic patterns into sounds, with peak height modulating volume and feature width into frequency. This mixture of visual and sound produces a unique sound pattern for each fingerprint. The initial scan point and scan speed are selectable, so a single fingerprint can produce different sets of sounds. In addition to being a playful app, this work can be used to aid those working with fingerprint analysis to provide additional selection criteria.
A couple of researchers from China demonstrated image modification work. Chen Cao showed off another version of facial recognition and animation tools. With a simple web-cam, he can do real-time capture and recognition, which can be applied to the display. The algorithms are fast enough to allow for real-time motion capture and apply them to an avatar. The camera has to have sufficient resolution to see fairly fine detail, and the code size is small enough to run on a smart phone. The challenge is the time to acquire the training data, and the variability in varying light.
His colleague presented a 3-d aware hair manipulator for single images. This tool allows the user to structure and shape hair on a model. The input is a 2-D image that the user can manipulate in place, or can replace completely with hair from a file. This tools lets the user view styling changes and appearance editing as well as a complete hair replacement from any other source.
The portrait is transformed into a 3-D image for full turntable viewing. This dynamic application simulates many hair effects and is a real-time video manipulator. It’s available as an iPhone or iPad app. The capture function is quick, and the hair definition just requires an outline of the hair. The user can change the hair or other parts and then can move the face while the hair stays in proper position on the head, even while moving.
