Log Focal Frequency Loss for Bioimage Restoration
Xingjian Zhang, Claire Leclech, Louison Blivet-Bailly, Abdul I. Barakat, Elsa D. Angelini
TL;DR
Biological fluorescence microscopy restoration suffers from spectral bias and foreground-background imbalance, hindering recovery of fine textures and edges. The authors introduce Log Focal Frequency Loss ($LFFL$), which combines adaptive log-space weighting with a log-dampened error in the Fourier domain and integrates this loss into a GAN-based restoration framework. Across deblurring and denoising tasks on real ground-truth datasets, $LFFL$ yields consistent gains over spatial losses and prior frequency-domain losses, with a tunable focal parameter $\alpha$ that trades perceptual quality against pixel fidelity. The approach is lightweight to implement within existing GAN pipelines and offers practical improvements for bioimage restoration in microscopy, enabling better preservation of high-frequency details without overwhelming gradients by large low-frequency amplitudes.
Abstract
Image restoration of biological structures in microscopy poses unique challenges for preserving fine textures and sharp edges. While recent GAN-based image restoration formulations have introduced frequency-domain losses for natural images, microscopy images pose distinct challenges with large dynamic ranges and sparse but critical structures with spatially-variable contrast. Inspired by the principle of logarithmic perception in human vision, we propose a log focal frequency loss (LFFL) tailored for microscopy restoration. This loss combines adaptive spectral weighting from log-space differences with log-dampened error measurement, ensuring balanced reconstruction across all frequency bands while preserving both structural coherence and fine details. We tested our GAN-based framework on two use-cases with real ground-truths: deblurring of fluorescence images of cell nuclei on microgroove substrates and denoising of zebrafish embryo images from the FMD dataset. Compared to training with only spatial-domain losses and with existing frequency-domain losses, our method achieves improvements across several quality metrics. Code is available at github.com/xjzhaang/log-focal-frequency-loss.