Depth and DOF Cues Make A Better Defocus Blur Detector

Defocus blur detection (DBD) separates in-focus and out-of-focus regions inan image. Previous approaches mistakenly mistook homogeneous areas in focus fordefocus blur regions, likely due to not considering the internal factors thatcause defocus blur. Inspired by the law of depth, depth of field (DOF), anddefocus, we propose an approach called D-DFFNet, which incorporates depth andDOF cues in an implicit manner. This allows the model to understand the defocusphenomenon in a more natural way. Our method proposes a depth featuredistillation strategy to obtain depth knowledge from a pre-trained monoculardepth estimation model and uses a DOF-edge loss to understand the relationshipbetween DOF and depth. Our approach outperforms state-of-the-art methods onpublic benchmarks and a newly collected large benchmark dataset, EBD. Sourcecodes and EBD dataset are available at: https:github.com/yuxinjin-whu/D-DFFNet.