Inter-Image Pixel Shuffling for Multi-focus Image Fusion

Huangxing Lin; Rongrong Ma; Cheng Wang

cs.CVMar 7, 2026

Inter-Image Pixel Shuffling for Multi-focus Image Fusion

Authors: Huangxing Lin, Rongrong Ma, Cheng Wang

Organizations: College of Computer Science and Technology, Huaqiao University, Xiamen 361021, China

Abstract

Multi-focus image fusion aims to combine multiple partially focused images into a single all-in-focus image. Although deep learning has shown promise in this task, its effectiveness is often limited by the scarcity of suitable training data. This paper introduces Inter-image Pixel Shuffling (IPS), a novel method that allows neural networks to learn multi-focus image fusion without requiring actual multi-focus images. IPS reformulates the task as a pixel-wise classification problem, where the goal is to identify the focused pixel from a pixel group at each spatial position. In this method, pixels from a clear optical image are treated as focused, while pixels from a low-pass filtered version of the same image are considered defocused. By randomly shuffling the focused and defocused pixels at identical spatial positions in the original and filtered images, IPS generates training data that preserves spatial structure while mixing focus-defocus information. The model is trained to select the focused pixel from each spatially aligned pixel group, thus learning to reconstruct an all-in-focus image by aggregating sharp content from the input. To further enhance fusion quality, IPS adopts a cross-image fusion network that integrates the localized representation power of convolutional neural networks with the long-range modeling capabilities of state space models. This design effectively leverages both spatial detail and contextual information to produce high-quality fused results. Experimental results indicate that IPS significantly outperforms existing multi-focus image fusion methods, even without training on multi-focus images.

Figures & tables

Fig. 1: Cartoon schematic of multi-focus image fusion. Red boxes indicate focused pixels, while blue boxes represent defocused pixels.

Fig. 2: Illustration of pixel shuffling. Red boxes indicate focused pixels, while blue boxes represent defocused pixels. (a) Applying pixel shuffling to the multi-focus images yields a pair of recombined images. Since both the multi-focus and recombined image pairs share the same pixels, they correspond to the same all-in-focus image. (b) Filtering and pixel shuffling a single optical image can produce recombined images identical to those in (a). These recombined images can thus serve as substitutes for learning multi-focus image fusion, with the optical image in (b) used as the ground-truth.

Fig. 3: IPS training pipeline. Red boxes indicate focused pixels, while blue boxes represent defocused pixels. The mask $m$ shares the same dimensions as the image.

Fig. 4: Architecture of the Cross-Image Fusion Network.

Methods	$\mathcal{Q}_{MI}$	$\mathcal{Q}_{SF}$	$\mathcal{Q}_{S}$	$\mathcal{Q}_{CB}$	$\mathcal{Q}_{AB/F}$	$\mathcal{Q}_{NCIE}$
DSIFT [ 37 ]	1.0271	0.0685	0.8158	0.7200	0.6760	0.8194
DCT_Corr [ 38 ]	1.0197	0.0674	0.8173	0.7103	0.6419	0.8191
CNN [ 14 ]	1.0124	0.0748	0.8373	0.7878	0.7260	0.8325
IFCNN [ 39 ]	1.0325	0.0745	0.8253	0.7169	0.7206	0.8270
SwinFusion [ 40 ]	1.0217	0.0735	0.8302	0.7218	0.7140	0.8237
U2Fusion [ 41 ]	0.7027	0.0721	0.7213	0.5533	0.6089	0.8191

TABLE I: Quantitative results on the Lytro dataset. Best results are shown in red and the second-best in blue .

Methods	$\mathcal{Q}_{MI}$	$\mathcal{Q}_{SF}$	$\mathcal{Q}_{S}$	$\mathcal{Q}_{CB}$	$\mathcal{Q}_{AB/F}$	$\mathcal{Q}_{NCIE}$
DSIFT [ 37 ]	0.6902	0.0835	0.6460	0.6210	0.6064	0.8168
DCT_Corr [ 38 ]	0.6897	0.0838	0.6547	0.6087	0.5958	0.8168
CNN [ 14 ]	0.7152	0.0833	0.6423	0.6207	0.6048	0.8175
IFCNN [ 39 ]	0.7003	0.0853	0.6635	0.5882	0.5910	0.8170
SwinFusion [ 40 ]	0.7001	0.0621	0.6333	0.5650	0.5834	0.8167
U2Fusion [ 41 ]	0.6436	0.0879	0.5747	0.5183	0.5281	0.8148

TABLE II: Quantitative results on the MFFW dataset. Best results are shown in red and the second-best in blue .

Methods	Real-MFF		MFI-WHU
Methods	PSNR	SSIM	PSNR	SSIM
DSIFT [ 37 ]	39.94	0.987	45.26	0.992
DCT_corr [ 38 ]	30.65	0.982	42.97	0.990
CNN [ 14 ]	40.55	0.988	45.05	0.993
IFCNN [ 39 ]	40.77	0.990	40.61	0.993
SwinFusion [ 40 ]	28.09	0.978	31.51	0.951

TABLE III: Quantitative results on the Real-MFF and MFI-WHU datasets. Best results are shown in red and the second-best in blue .

Fig. 5: Visual comparison of fused images on the Lytro dataset. Corresponding difference maps with respect to Source Image 1 are also presented.

Fig. 6: Visual comparison of fused images on the MFFW dataset. Corresponding difference maps with respect to Source Image 1 are also presented.

Fig. 7: Visual comparison of fusion results on the Real-MFF dataset. Difference maps with respect to the ground-truth all-in-focus images are provided for reference.

Fig. 8: Visual comparison of fusion results on the MFI-WHU dataset. Difference maps with respect to the ground-truth all-in-focus images are provided for reference.

Network	PSNR	SSIM
w/o global branch	36.98	0.965
w/o local branch	38.83	0.986
Both branches retained	42.19	0.991

TABLE IV: Quantitative results of the ablation study on network architecture. “w/o” denotes “without”.

Fig. 9: Visual results of network architecture ablation study.

Filter	PSNR	SSIM
Median filter	41.51	0.987
Gaussian filter	41.86	0.989
Mean filter	42.19	0.991

TABLE V: Quantitative comparison of IPS fusion performance using different filters.

Mask ratio	PSNR	SSIM
$p=0.1$	36.19	0.962
$p=0.3$	36.26	0.963
$p=0.5$	42.19	0.991
$p=0.7$	36.28	0.963
$p=0.9$	36.14	0.962

TABLE VI: Quantitative comparison of IPS fusion performance using different mask ratios.

Explore similar work

Aug 4, 2026

Clarity Contrast and Similarity Selection for Multi-Focus Image Fusion

Multi-focus image fusion (MFIF) aims to generate an all-in-focus image from multiple images of the same scene focused at different regions. Most existing deep learning-based methods lack explicit interaction between the source images, which limits their performance and interpretability. This paper presents a novel Clarity Contrast and Similarity Selection Network (CSNet), to bridge direct information exchange for MFIF. Specifically, by contrasting the clarity differences between source images within our proposed Clarity Contrast Attention Module (CCAM), we mutually enhance sharp features while suppressing blurry ones. This allows us to identify the exactly focused regions in each source and locate the focused-defocused boundaries. Moreover, the Defocus Spread Effect (DSE) degrades pixels in all source images around the boundaries. To further refine these ambiguous areas, we introduce a Similarity Selection Strategy, which reconstructs an initial clear image from source images and selects optimal pixels by comparing the similarity among them. Through this interactive approach, CSNet effectively preserves focused regions as well as recovering natural boundaries to fuse an all-in-focus output. Extensive experiments demonstrate that our method achieves state-of-the-art performance both quantitatively and qualitatively. Our code is available on Github: https://github.com/ZYC-HUST/CSNet.

Yicheng Zhang, Haoyou Deng, Zhiqiang Li +3

Key Laboratory of Image Processing and Intelligent Control, School of Artificial Intelligence and Automation, Huazhong University of Science and Technology, Wuhan, 430074, China

Multi-Source Image Fusion

Ask PDF

Jun 30, 2026

UHD-MFF: Shattering Barriers in Multi-Focus Ultra-High-Definition Image Fusion via Learnable Lookup Tables

With the advancement of imaging technology, ultra-high-definition images have become increasingly essential in modern visual applications. However, existing multi-focus image fusion remains largely confined to low-resolution images and faces three major barriers in UHD scenarios, namely data availability, model adaptability, and deployment feasibility, which severely hinder its practical application. To shatter these barriers, first, we propose the UHD-MFF dataset, the first large-scale ultra-high-resolution multi-focus fusion dataset. Second, we propose a scale-specialized lookup-table framework tailored for ultra-high-resolution images, termed as UMF-LUT. It consists of Coarse-Region Lookup Table (C-LUT) and Detail-Edge Lookup Table (D-LUT). Specifically, C-LUT performs joint queries of multiple gradient cues and semantic cues at low-resolution scales to enable region-level decision-making. Also, D-LUT operates at high-resolution scales, leveraging efficient Laplacian cues to provide complementary edge-level decision information. Such a design makes the model particularly well-suited for ultra-high-resolution multi-focus image fusion. Finally, it offers strong deployability with minimal computational overhead, enabling real-time 4K multi-focus fusion and showing promising potential for smartphone. Extensive experiments demonstrate that it outperforms SOTA methods in both visual fidelity and quantitative metrics. It effectively advances the development of multi-focus image fusion toward ultra-high-resolution imaging scenarios. The code is available at https://github.com/zyb5/UHD-MFF.

Yibing Zhang, Xunpeng Yi, Qinglong Yan +3

Electronic Information School, Wuhan University, Wuhan 430072, China · School of Automation, Southeast University, Nanjing 210096, China · School of Robotics, Wuhan University, Wuhan 430072, China

Multi-Source Image Fusion

Ask PDF

Mar 15, 2026

UniFusion: A Unified Image Fusion Framework with Robust Representation and Source-Aware Preservation

Image fusion aims to integrate complementary information from multiple source images to produce a more informative and visually consistent representation, benefiting both human perception and downstream vision tasks. Despite recent progress, most existing fusion methods are designed for specific tasks (i.e., multi-modal, multi-exposure, or multi-focus fusion) and struggle to effectively preserve source information during the fusion process. This limitation primarily arises from task-specific architectures and the degradation of source information caused by deep-layer propagation. To overcome these issues, we propose UniFusion, a unified image fusion framework designed to achieve cross-task generalization. First, leveraging DINOv3 for modality-consistent feature extraction, UniFusion establishes a shared semantic space for diverse inputs. Second, to preserve the understanding of each source image, we introduce a reconstruction-alignment loss to maintain consistency between fused outputs and inputs. Finally, we employ a bilevel optimization strategy to decouple and jointly optimize reconstruction and fusion objectives, effectively balancing their coupling relationship and ensuring smooth convergence. Extensive experiments across multiple fusion tasks demonstrate UniFusion's superior visual quality, generalization ability, and adaptability to real-world scenarios. Code is available at https://github.com/dusongcheng/UniFusion.

Xingyuan Li, Songcheng Du, Yang Zou +3

Zhejiang University · Northwest Polytechnical University · Dalian University of Technology +1

Multi-Source Image Fusion

Ask PDF