Search papers, labs, and topics across Lattice.
This paper introduces Bio-SFT, a bio-inspired spiking frequency transformer designed for single-image high dynamic range (HDR) reconstruction from standard dynamic range (SDR) images. By integrating a learnable retinal adaptation frontend, an asymmetric Parvo-to-Magno guidance mechanism, and an event-driven spiking neural network (SNN) gating module, Bio-SFT effectively stabilizes input under varying lighting conditions and reduces noise in dark regions. Experimental results on HDRTV1K demonstrate that Bio-SFT not only enhances perceptual quality but also significantly reduces artifact propagation compared to traditional symmetric guidance methods.
Bio-SFT achieves superior HDR reconstruction by leveraging biological principles, outperforming conventional methods in both quality and efficiency.
Recovering high dynamic range (HDR) radiance from a single standard dynamic range (SDR) image is highly ill-posed. Extreme luminance variation and severe quantization in dark regions make accurate reconstruction challenging, often leading to visual artifacts and color distortions. To address this problem, we propose Bio-SFT, a bio-inspired spiking frequency transformer for single-image HDR reconstruction. Bio-SFT incorporates three biologically motivated components. First, a learnable Naka--Rushton retinal adaptation frontend stabilizes the input under complex lighting conditions. Second, an explicit Parvo--Magno split introduces asymmetric Parvo-to-Magno guidance, allowing high-frequency structural cues to modulate low-frequency reconstruction. Third, an event-driven SNN hard gating module applies all-or-none spiking to suppress dark-region noise while preserving structural details. The module is trained with a sparsity prior to encourage efficient feature utilization. Built for end-to-end training within a transformer backbone, these lightweight components provide strong parameter efficiency. Experiments on HDRTV1K show that Bio-SFT achieves competitive perceptual quality and consistently improves HDR-VDP-3 and $螖E_{ITP}$ while reducing artifact propagation in symmetric guidance pipelines.
