AI summaryⓘ
The authors studied a problem called flicker-banding that happens when a camera records a screen, making the picture look strange with color shifts and stripes. They found that these stripes change a lot depending on camera settings, so they developed a new method using multiple photos taken at different exposures to better fix the issue. They created a special dataset combining computer simulations and real photos to train their model, called BRACE, which smartly combines information from multiple images to remove the stripe artifacts. They also introduced a way to measure how well the stripes are removed. Tests showed their method works better than previous ones on both fake and real images.
flicker-bandingrolling shuttertemporal aliasingmulti-exposure imagingRAW image formatray tracingmulti-frame restorationfrequency-aware processingspatial cross-attentionimage artifact removal
Authors
Zihan Zhou, Libo Zhu, Jue Gong, Zhiyi Zhou, Jiezhang Cao, Yong Guo, Yulun Zhang
Abstract
Flicker-banding (FB), arises from temporal aliasing between a camera's rolling shutter and a display's brightness modulation, degrading screen-captured image readability with color shifts and jagged patterns. Existing single-frame methods with simplified parametric stripe models cannot reliably distinguish these artifacts from genuine texture. To address this, we conduct a systematic analysis of complex FB morphologies and reveal their significant variation across exposure settings, motivating a multi-frame bracketed RAW restoration paradigm. We construct Bricker, a synthetic-real bracketed RAW dataset built via ray-tracing-based physical simulation and automated multi-exposure capture tool. We further propose BRACE: Bracketed RAW Flicker-Banding Removal, a multi-frame restoration model that utilizes frequency-aware banding prior and a multi-scale spatial cross-attention modulator (MSCAM) for cross-exposure spatial fusion. We also introduce the Stripe Frequency Consistency (SFC) metric to evaluate banding removal. Experiments demonstrate state-of-the-art performance on both synthetic and real benchmarks. Our dataset and code are available at: https://github.com/ZZH-qwq/BRACE.