When images and videos are not aesthetically pleasing, human viewers stop in the middle and go elsewhere on the web for a better viewing experience. One main reason is the transitions between objects (edges) are blurred or not distinct. Possible solutions include some techniques to smooth out the shades so that the objects can stand out better. They are discussed in the author’s dissertation on efficient high-dimensional edge-aware filtering.

The dissertation focuses on two new filtering approaches: “the domain transform for geodesic response and the adaptive manifolds for Euclidean response.” With these frameworks, several edge-aware filters are proposed to provide the fastest performance (on both central processing units (CPUs) and graphics processing units (GPUs)) for various real-world applications.

The edge-aware filtering approaches have caught the attention of the image and video processing, computer graphics, computer vision, and computational photography communities. According to the paper, “third-party implementations of the domain transform and adaptive manifolds have been included in the open-source computer vision library (OpenCV).”

A recipient of the 2016 ACM SIGGRAPH Outstanding Doctoral Dissertation Award, the author keeps the presentation informal with illustrations and examples to show the differences between blurred and more distinct transitions between the edges. Those interested in the challenges of improving edge-aware filtering approaches should read this paper.