Perceptual Preferences Applied To Massive Super-Resolution and Image Editing
The current state of the art in image super-resolution relies on interpolation and patch-based approaches and stops at a few orders of magnitude. This project poses the pertinent question of what lies between the pixels of an image and how we could infer or "hallucinate" appropriate content to fill those spaces at much greater orders of magnitude.
A key component of the work shall investigate preferences of many human observers to crowd source a model of the details people would expect to appear when massively up-scaling an image. Such human-sourced image understanding may facilitate other novel image editing operations.
This project will rely on merging cross disciplinary knowledge from art, human perceptual psychology, computer vision and machine learning.
Other projects that I am currently working on are:
A personal research project in which I am investigating aspects of perceived image aesthetics. Such models have great applications within the image manipulation and editing domain. A group project focused on creating fully functional articulated models for 3D printing.