Frequency-Driven Inverse Kernel Prediction for Single Image Defocus Deblurring

TL;DR

This paper introduces a novel frequency-domain based neural network for single image defocus deblurring, improving kernel estimation accuracy and deblurring quality through dual-branch prediction, adaptive convolution, and scale recurrent modules.

Contribution

It proposes a frequency-driven inverse kernel prediction network with dual-branch strategy, position adaptive convolution, and dual-domain scale recurrent modules for superior defocus deblurring.

Findings

Outperforms existing defocus deblurring methods in experiments

Effective frequency domain representations enhance kernel modeling

Improved deblurring quality from coarse to fine levels

Abstract

Single image defocus deblurring aims to recover an all-in-focus image from a defocus counterpart, where accurately modeling spatially varying blur kernels remains a key challenge. Most existing methods rely on spatial features for kernel estimation, but their performance degrades in severely blurry regions where local high-frequency details are missing. To address this, we propose a Frequency-Driven Inverse Kernel Prediction network (FDIKP) that incorporates frequency-domain representations to enhance structural identifiability in kernel modeling. Given the superior discriminative capability of the frequency domain for blur modeling, we design a Dual-Branch Inverse Kernel Prediction (DIKP) strategy that improves the accuracy of kernel estimation while maintaining stability. Moreover, considering the limited number of predicted inverse kernels, we introduce a Position Adaptive Convolution (PAC) to enhance the adaptability of the deconvolution process. Finally, we propose a Dual-Domain Scale Recurrent Module (DSRM) to fuse deconvolution results and progressively improve deblurring quality from coarse to fine. Extensive experiments demonstrate that our method outperforms existing approaches. Code will be made publicly available.