Computational imaging (CI) is a discipline at the intersection of computer vision, image processing, optics, and AI. Its goal is to expand the capabilities of conventional imaging by combining novel imaging techniques with advanced algorithms. Modern computational imaging techniques power every smartphone camera—from stunning portraits to low-light photography —they are, in fact, what make smartphone cameras work. Computational imaging research, however, strives to go beyond the capabilities of conventional imaging altogether, yielding "cameras" that can capture videos at the speed of light, see around corners or below the skin, and cameras that capture speech and music directly from the vibration they create on object surfaces.
This course lays the foundations of computational imaging while also covering some state-of-the-art research. Tentative topics include:
- The modern image processing pipelines, basic optics, the image formation model (lenses, aberrations, sensor noise, and imaging in color), and computational light transport.
- Advanced image and video editing algorithms like filtering, gradient-domain processing, and deconvolution.
- Advanced image acquisition techniques like light-field imaging, coded photography, focal stacks, depth from defocus, time-of-flight imaging, 3D scanning, and more.
This course involves hands-on experience through multiple homework assignments, where you will get to implement various foundational computational imaging algorithms covered in class. The final project consists in creating a short 5-minute (CVPR-style) video describing a hot recent CI paper.