BOCCHI: A More Realistic and Challenging Benchmark for Local Motion Blur Detection with MSDCT-UNet
2026-07-11 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors created a new dataset called BOCCHI for detecting blurry areas in photos, which is more realistic and prevents models from cheating by using easy shortcuts. They also developed a new model called MSDCT-UNet that uses special frequency-based techniques to better identify blur at different scales. Their model performed best on BOCCHI and was better at generalizing to other datasets when trained on BOCCHI data. This work shows improved accuracy in finding blurry regions in images using real-world data and frequency information.
motion blur detectionpixel-level localizationbenchmark datasetBocchi datasetMulti-Scale Discrete Cosine Transform (DCT)UNet architectureDCT AttentionFiLM layerscross-dataset transfermean Intersection over Union (mIoU)
Authors
Kuan-Lin Chen, Yuan-Kang Lee, Cheng-Yuan Chiang, Jian-Jiun Ding
Abstract
Local motion blur detection requires pixel-level localization of blurred regions. Existing benchmarks let models rely on gradient shortcuts that fail to transfer. We introduce BOCCHI (Blurred Objects Captured across Cameras with Human-annotated Imagery), a real-captured benchmark whose sharp regions overlap the blur gradient distribution and defeat these shortcuts, and propose MSDCT-UNet (Multi-Scale Discrete Cosine Transform UNet), a frequency-aware encoder-decoder injecting multi-scale DCT priors through DCT Attention and FiLM. MSDCT-UNet ranks first in in-domain mIoU and boundary localization on BOCCHI, and BOCCHI-trained models outperform every other training source on cross-dataset transfer with only 633 training images.