Efficient Inference in Fully Connected CRFs with Gaussian Edge Potentials
This work addresses a computational bottleneck for researchers and practitioners in computer vision, enabling dense pixel-level connectivity in CRFs to enhance segmentation tasks, though it is incremental as it builds on existing CRF frameworks.
The paper tackled the problem of performing inference in fully connected conditional random fields (CRFs) for image segmentation, which are computationally prohibitive due to billions of edges, by developing an efficient approximate inference algorithm that uses Gaussian kernels, resulting in substantial improvements in segmentation and labeling accuracy.
Most state-of-the-art techniques for multi-class image segmentation and labeling use conditional random fields defined over pixels or image regions. While region-level models often feature dense pairwise connectivity, pixel-level models are considerably larger and have only permitted sparse graph structures. In this paper, we consider fully connected CRF models defined on the complete set of pixels in an image. The resulting graphs have billions of edges, making traditional inference algorithms impractical. Our main contribution is a highly efficient approximate inference algorithm for fully connected CRF models in which the pairwise edge potentials are defined by a linear combination of Gaussian kernels. Our experiments demonstrate that dense connectivity at the pixel level substantially improves segmentation and labeling accuracy.