PCDFusion: Proposal-Context-Detail Bayesian Rendering for Infrared-Visible Image Fusion

Rui Liu; Lin Gu; Hesong Li; Ying Fu

cs.CVOct 16, 2026

PCDFusion: Proposal-Context-Detail Bayesian Rendering for Infrared-Visible Image Fusion

Authors: Rui Liu, Lin Gu, Hesong Li, Ying Fu

Organizations: School of Computer Science and Technology, Beijing Institute of Technology, Beijing 100081, China · Research Institute of Electrical Communication, Tohoku University, Sendai 980-8577, Japan

Abstract

Infrared-visible image fusion (IVIF) aims to combine visible and thermal infrared observations for human perception and downstream vision tasks. Visible and thermal infrared observations are physically complementary but unreliable in different ways. Existing deep IVIF methods mainly improve feature interaction and reconstruction, but often lack an explicit local rule for deciding when visible appearance should dominate and when thermal responses should contribute. To address this issue, we reformulate IVIF as local Bayesian posterior rendering, where fusion is controlled by posterior action weights rather than direct modality mixing. In this paper, we propose Proposal-Context-Detail Fusion (PCDFusion), a visible-anchored framework that implements this formulation through three rendering stages. PCDFusion preserves potentially useful thermal responses, updates exposure-degraded visible regions, and renders reliable band-limited infrared structures into the final luminance. Extensive experiments on public IVIF benchmarks show that PCDFusion improves fusion quality, target recovery, and downstream detection and segmentation performance under challenging illumination. The code is available at https://github.com/BIT-RuiLiu/PCDFusion.git.

Figures & tables

Figure 1: PCDFusion estimates posterior action weights from physical cues and cross-modal observations, recovering structures under visible failure while suppressing unreliable thermal responses.

Figure 2: Overall architecture of the proposed PCDFusion.

Panel of Table 1 — Table 1: Quantitative fusion results on the MSRS, FMB, and M3FD datasets. The best results are highlighted in bold , and the second-best results are underlined .

Panel of Table 2 — Table 2: Downstream perception results on FMB semantic segmentation and M3FD object detection.

Table 5

Figure 5: Left: luminance and RGB outputs of the proposal, context, and detail rendering stages. Right: priors, learned log-evidence scores, and posterior action weights of the three Bayesian processes.

Panel of Figure 6 — Figure 6: Calibration of the exposure-failure prior and posterior context update weight.

Table 8

Table 9

Appendix figures & tables19 assets

Supplementary material from the paper’s appendix.

Appendix

Role	Probabilistic operation	PCDFusion realization	Consequence
Decision	Bernoulli rendering action $z_{k}$	Explicit proposal/context/detail action semantics	Defined local inference target
Prior	$p(z_{k}=1\mid\mathcal{P}_{k})$	Analytic IR-response, exposure, or structure cue	Physical zero-evidence reference
Evidence	Relative support in log Bayes-factor coordinates	Neural score $f_{\theta_{k}}(\mathcal{O}_{k},\mathcal{P}_{k})$	Amortized log-evidence correction
Update	Prior odds $\times$ evidence factor	$\mathbf{q}_{k}=\sigma(\operatorname{logit}(\boldsymbol{\pi}_{k})+\boldsymbol{\eta}_{k})$	Normalized action posterior
Rendering	Expectation over two actions	$(1-\mathbf{q}_{k})\mathcal{R}_{k,0}+\mathbf{q}_{k}\mathcal{R}_{k,1}$	Posterior-weighted fusion
Learning	Shared estimator of observation evidence	End-to-end optimization of $\theta_{k}$ through rendering losses	Amortized across pixels and images

Appendix

Table 9: Probabilistic roles and their realizations in PCDFusion.

Asset	Type	Repository / Source	License / Terms
SwinFusion JAS’22	Baseline implementation	https://github.com/Linfeng-Tang/SwinFusion.git	Not specified
TarDAL CVPR’22	Baseline implementation / M3FD source	https://github.com/JinyuanLiu-CV/TarDAL.git	GPL-3.0
CDDFuse CVPR’23	Baseline implementation	https://github.com/Zhaozixiang1228/MMIF-CDDFuse.git	Not specified
Diff-IF Inf. Fusion’24	Baseline implementation	https://github.com/XunpengYi/Diff-IF.git	MIT
EMMA CVPR’24	Baseline implementation	https://github.com/Zhaozixiang1228/MMIF-EMMA.git	Not specified
DCEvo CVPR’25	Baseline implementation	https://github.com/Beate-Suy-Zhang/DCEvo.git	Not specified

Appendix

Table 10: External assets used in this work. “Not specified” means that no explicit repository-level license is shown on the public repository page.

Method	EN $\uparrow$	AG $\uparrow$	MI $\uparrow$	SD $\uparrow$	SCD $\uparrow$	VIF $\uparrow$	Qabf $\uparrow$	SSIM $\uparrow$	SF $\uparrow$	CE $\downarrow$
SwinFusion JAS’22	6.620	3.562	4.393	42.980	1.686	1.032	0.647	0.721	11.096	0.052562
TarDAL CVPR’22	6.211	2.926	2.669	32.781	1.375	0.716	0.416	0.540	9.547	0.065997
CDDFuse CVPR’23	6.699	3.749	4.705	43.368	1.618	1.084	0.676	0.709	11.567	0.053406
Diff-IF Inf. Fusion’24	6.669	3.696	4.970	42.596	1.624	1.097	0.685	0.711	11.460	0.054479
EMMA CVPR’24	6.715	3.786	4.129	44.568	1.628	1.016	0.632	0.711	11.574	0.052032
DCEvo CVPR’25	6.622	3.725	4.157	42.926	1.662	1.066	0.688	0.721	11.470	0.050816

Appendix

Table 11: Complete image-level fusion results on MSRS. The best results are highlighted in bold , and the second-best results are underlined .

Method	EN $\uparrow$	AG $\uparrow$	MI $\uparrow$	SD $\uparrow$	SCD $\uparrow$	VIF $\uparrow$	Qabf $\uparrow$	SSIM $\uparrow$	SF $\uparrow$	CE $\downarrow$
SwinFusion JAS’22	6.683	4.052	4.270	35.352	1.560	1.032	0.665	0.743	13.515	0.000238
TarDAL CVPR’22	6.963	3.541	3.423	40.316	1.573	0.807	0.462	0.745	11.718	0.000369
CDDFuse CVPR’23	6.778	4.306	4.148	37.043	1.622	1.061	0.672	0.745	14.577	0.000230
Diff-IF Inf. Fusion’24	6.632	4.047	4.985	34.191	1.368	1.047	0.660	0.735	13.869	0.000229
EMMA CVPR’24	6.775	4.684	3.947	36.805	1.504	1.031	0.645	0.735	15.006	0.000224
DCEvo CVPR’25	6.673	4.051	4.512	34.749	1.503	1.058	0.686	0.747	13.771	0.000218

Appendix

Table 12: Complete image-level fusion results on FMB. The best results are highlighted in bold , and the second-best results are underlined .

Method	EN $\uparrow$	AG $\uparrow$	MI $\uparrow$	SD $\uparrow$	SCD $\uparrow$	VIF $\uparrow$	Qabf $\uparrow$	SSIM $\uparrow$	SF $\uparrow$	CE $\downarrow$
SwinFusion JAS’22	6.808	4.606	4.020	35.776	1.560	0.898	0.607	0.717	13.672	0.031855
TarDAL CVPR’22	6.985	4.265	3.130	38.785	1.523	0.708	0.427	0.708	12.672	0.029532
CDDFuse CVPR’23	6.907	4.861	3.774	37.206	1.646	0.917	0.610	0.718	14.753	0.031766
Diff-IF Inf. Fusion’24	6.791	4.567	4.852	34.927	1.322	0.894	0.589	0.701	14.115	0.031833
EMMA CVPR’24	6.927	5.335	3.753	38.253	1.494	0.897	0.590	0.703	15.235	0.031770
DCEvo CVPR’25	6.842	4.620	4.298	35.582	1.491	0.917	0.633	0.720	14.109	0.031699

Appendix

Table 13: Complete image-level fusion results on M3FD. The best results are highlighted in bold , and the second-best results are underlined .

Method	Road	Side.	Build.	Light	Sign	Veg.	Sky	Per.	Car	Truck	Bus	Motor.	Pole	mIoU $\uparrow$
SwinFusion JAS’22	89.22	48.98	83.16	43.22	80.68	87.16	94.80	70.58	84.23	48.77	57.57	36.28	50.23	58.33
TarDAL CVPR’22	89.50	50.05	83.24	42.30	81.20	88.21	95.00	73.31	81.85	35.82	26.36	37.18	49.76	55.59
CDDFuse CVPR’23	90.18	50.97	84.38	40.85	81.80	88.40	95.16	72.22	82.90	22.14	55.70	33.60	51.56	56.66
Diff-IF Inf. Fusion’24	89.66	43.81	83.64	45.24	82.04	87.02	94.70	70.69	83.57	45.22	75.99	32.26	50.82	58.98
EMMA CVPR’24	89.61	49.32	84.57	49.42	81.24	88.15	95.29	70.75	82.58	42.99	59.41	32.61	51.07	58.47
DCEvo CVPR’25	89.72	48.01	84.39	46.98	81.01	88.14	95.27	71.50	84.55	43.67	49.08	33.89	49.73	57.73

Appendix

Table 14: Complete FMB semantic segmentation results. The zero-valued classes are omitted. The best results are highlighted in bold , and the second-best results are underlined .

Method	mAP 50							mAP 50:95
Method	Per.	Car	Bus	Motor.	Lamp	Truck	mAP	Per.	Car	Bus	Motor.	Lamp	Truck	mAP
SwinFusion JAS’22	82.28	91.25	89.10	66.22	79.83	78.75	81.24	48.90	68.13	71.31	40.82	42.03	58.00	54.86
TarDAL CVPR’22	81.94	90.88	87.43	59.35	78.67	76.25	79.09	48.80	66.51	69.55	34.00	38.87	56.38	52.35
CDDFuse CVPR’23	81.93	91.57	90.26	68.36	81.59	74.18	81.31	48.79	68.44	71.80	38.00	43.44	54.39	54.15
Diff-IF Inf. Fusion’24	82.81	91.38	88.46	66.32	77.91	79.14	81.00	49.77	67.43	70.08	38.47	39.82	58.68	54.04
EMMA CVPR’24	81.70	91.11	90.46	63.35	81.59	77.36	80.93	48.35	68.03	71.02	38.26	43.70	57.20	54.43

Appendix

Table 15: Complete M3FD object detection results. The best results are highlighted in bold , and the second-best results are underlined .

Metric	SwinFusion	TarDAL	CDDFuse	Diff-IF	EMMA	DCEvo	Deno-IF	SAGE	UPFusion	PCDFusion
Over-exposed IR-edge retention (%) $\uparrow$	16.8	27.0	51.3	47.4	24.0	62.9	49.0	25.7	17.3	66.3
Reliable-sky preservation (%) $\uparrow$	100.0	0.1	99.5	99.6	97.5	99.5	7.3	0.0	99.9	99.5

Appendix

Table 16: Local structure recovery and appearance preservation in the first MSRS visualization. The best result is highlighted in bold , and the second-best result is underlined .

Fusion method	FMB mIoU (%)			M3FD mAP 50:95 (%)
Fusion method	VIS-trained	IR-trained	Mean	VIS-trained	IR-trained	Mean
SwinFusion	54.18	40.39	47.28	25.96	8.46	17.21
TarDAL	51.76	42.46	47.11	24.48	10.04	17.26
CDDFuse	56.09	40.79	48.44	26.54	8.35	17.44
Diff-IF	53.61	40.01	46.81	23.30	10.34	16.82
EMMA	55.65	39.59	47.62	30.01	5.93	17.97

Appendix

Table 17: Single-modality-to-fused transfer results. VIS-trained and IR-trained task models are evaluated directly on fused test images; Mean averages the two scores. The best result is highlighted in bold , and the second-best result is underlined .

Method	Fusion level	Test-time inputs	Training protocol	Params (M) $\downarrow$	GFLOPs $\downarrow$	mAP 50:95 (%) $\uparrow$
PCDFusion + YOLOv10x	Pixel/image	One fused image	300 epochs	36.27	508.62	61.05
M2D-LIF	Feature	Paired VIS+IR	200-epoch teacher + 100-epoch student	36.51	122.42	61.26
LoME (width $0.90$ )	Feature	Paired VIS+IR	300 epochs	36.18	49.05	59.38

Appendix

Table 18: Capacity-matched comparison with feature-level RGB-IR detectors on M3FD. Parameter counts include the complete training architectures, and GFLOPs are measured for deployed

640\times 640

inference.

Figure 7: Additional visual comparisons on MSRS and FMB.

Figure 8: Additional FMB semantic segmentation and M3FD object detection visualizations.

Stage	Variant	AG $\uparrow$	SCD $\uparrow$	VIF $\uparrow$	Qabf $\uparrow$	Qcb $\uparrow$	mIoU $\uparrow$	mAP $\uparrow$
Proposal	Uniform: $0.5$	3.863	1.672	1.110	0.723	0.591	58.88	55.86
Proposal	Hard IR: $\mathbbm{1}[\operatorname{Norm}(\mathbf{Y}_{r})\geq 0.5]$	3.861	1.671	1.109	0.722	0.589	58.95	55.72
Proposal	Soft IR (Full): $\sigma(\tau_{p}(\operatorname{Norm}(\mathbf{Y}_{r})-0.5))$	3.839	1.671	1.117	0.723	0.593	60.05	55.98
Detail	Edge competition: $\mathbf{S}_{er}/(\mathbf{S}_{er}+\mathbf{S}_{v}+\epsilon)$	3.855	1.674	1.104	0.721	0.585	58.95	55.83
Detail	IR structure only: $\mathbf{S}_{r}$	3.848	1.677	1.110	0.722	0.587	58.84	55.70
Detail	IR intensity: $\operatorname{Norm}(\mathbf{Y}_{r})$	3.859	1.678	1.105	0.721	0.586	58.90	55.95

Appendix

Table 19: Additional ablation of proposal and detail prior variants. The best results are highlighted in bold , and the second-best results are underlined .

Parameter	Value	AG	VIF	Qabf	mIoU	mAP
Proposal temperature	$2$	3.813 ( $-0.026$ )	1.115 ( $-0.002$ )	0.720 ( $-0.003$ )	60.11 ( $+0.06$ )	55.94 ( $-0.05$ )
	$4$	3.832 ( $-0.007$ )	1.119 ( $+0.002$ )	0.723 ( $0.000$ )	60.10 ( $+0.05$ )	55.88 ( $-0.11$ )
	$6$ (default)	3.839 ( $0.000$ )	1.117 ( $0.000$ )	0.723 ( $0.000$ )	60.05 ( $0.00$ )	55.98 ( $0.00$ )
	$8$	3.830 ( $-0.009$ )	1.104 ( $-0.013$ )	0.720 ( $-0.002$ )	60.03 ( $-0.02$ )	55.90 ( $-0.08$ )
	$10$	3.794 ( $-0.045$ )	1.080 ( $-0.037$ )	0.714 ( $-0.009$ )	59.99 ( $-0.06$ )	55.95 ( $-0.04$ )
Detail scale	$0.050$	3.729 ( $-0.110$ )	1.109 ( $-0.008$ )	0.714 ( $-0.008$ )	59.90 ( $-0.15$ )	55.88 ( $-0.10$ )

Appendix

Table 20: Fixed-checkpoint parameter sensitivity. Parenthesized values show changes from the default setting.

Stage	Posterior-action map used for rendering	Local task gain $\uparrow$	FMB mIoU $\uparrow$
Proposal	Learned posterior	0.071	60.05
	Prior only	0.069	59.86
	Spatially shuffled	$-0.491$	50.21
	Inverted	0.003	58.68
	Zero map	0.000	58.58
Context	Learned posterior	0.013	60.05

Appendix

Table 21: Posterior-action interventions and local utility. Each intervention replaces one stage map while keeping the checkpoint and inputs fixed. Local gains and FMB mIoU are reported in percentage points.

Metric	SwinFusion JAS’22	TarDAL CVPR’22	CDDFuse CVPR’23	Diff-IF Inf. Fusion’24	EMMA CVPR’24	DCEvo CVPR’25	Deno-IF NIPS’25	SAGE CVPR’25	UPFusion AAAI’26	PCDFusion Ours
Params (M) $\downarrow$	0.97	0.30	1.19	23.74	1.52	2.00	1.43	0.14	154.82	4.60
GFLOPs $\downarrow$	604.42	155.07	945.04	6101.12	68.46	1561.87	221.15	34.11	2676.36	235.34
Time (ms) $\downarrow$	997.82	5.04	126.70	780.17	13.17	178.92	112.68	2.12	584.92	92.51
FPS $\uparrow$	1.00	198.55	7.89	1.28	75.92	5.59	8.87	471.34	1.71	10.81

Appendix

Table 22: Computational cost comparison on

512\times 512

inputs. The best results are highlighted in bold , and the second-best results are underlined .

Dataset	Estimate	Pixel failure			Class-normalized mFailure
Dataset	Estimate	Pearson $\uparrow$	Spearman $\uparrow$	MSE $\downarrow$	Pearson $\uparrow$	Spearman $\uparrow$	MSE $\downarrow$
MSRS	Darkness-only: $1-\mathbf{Y}_{v}$	0.762	0.982	0.1372	-0.111	-0.035	0.2231
	Brightness-only: $\mathbf{Y}_{v}$	-0.762	-0.982	0.3381	0.111	0.035	0.1936
	Distance-to-mid: $\left\|2\mathbf{Y}_{v}-1\right\|$	0.418	0.087	0.1825	0.815	0.928	0.0998
	Ours prior: $\left(2\mathbf{Y}_{v}-1\right)^{2}$	0.463	0.087	0.1014	0.879	0.928	0.0314
	Ours posterior: $\mathbf{q}_{c}$	0.646	0.147	0.0552	0.890	0.904	0.0658

Appendix

Table 23: Calibration of exposure-failure estimates on MSRS and FMB. The best results are highlighted in bold , and the second-best results are underlined .

Panel of Figure 9 — Figure 9: Relation between the analytic exposure-failure prior $U_{v}$ and empirical class-normalized visible failure on MSRS and FMB. Each point denotes a luminance bin, and Pearson/Spearman correlations are reported in the subplot titles.

Figure 10: Failure cases and limitations.

Infrared-Visible Image Fusion

Infrared-visible image fusion combines thermal cues with visible-light detail to create images for human viewing and machine perception.

See all 38 papers

P2Fusion: Prompt-based Progressive Infrared-Visible Image Fusion via Dual-Prior Distillation
Aug 13, 2026Yi Shi, Huichao Xie, Yuqing Wang +7Multi-Source Image Fusion Infrared-Visible Image Fusion
Northwestern Polytechnical University, Xi’an, China · Hefei University of Technology, Hefei, China · Rocket Force University of Engineering, Xi’an, China +1
EAPFusion: Intrinsic Evolving Auxiliary Prior Guidance for Infrared and Visible Image Fusion
May 3, 2026Zhenyu Sun, Luobin Zhang, Axi Niu +2Dynamic Convolution Multi-Source Image Fusion
Northwestern Polytechnical University, Xi’an, China
Bridging Human Evaluation to Infrared and Visible Image Fusion
Mar 4, 2026Jinyuan Liu, Xingyuan Li, Qingyun Mei +5Infrared-Visible Image Fusion Multi-Source Image Fusion
School of Mechanical Engineering, Dalian University of Technology · College of Computer Science and Technology, Zhejiang University · School of Software Technology & DUT-RU International School of ISE, Dalian University of Technology +1

Aug 13, 2026J/K move · Enter open · S save

P2Fusion: Prompt-based Progressive Infrared-Visible Image Fusion via Dual-Prior Distillation

Yi Shi, Huichao Xie, Yuqing Wang, Mingyu Wang+6

Northwestern Polytechnical University, Xi’an, China · Hefei University of Technology, Hefei, China · Rocket Force University of Engineering, Xi’an, China+1

Multi-Source Image Fusion Infrared-Visible Image Fusion Multimodal Fusion

Open paper AskPDF arXiv

Infrared-visible image fusion (IVIF) is pivotal for multimodal perception, yet reconciling the inherent information disparity between thermal and textural features remains a fundamental challenge. Existing prior-guided methods often rely on static constraints that induce optimization conflicts or utilize extrinsic semantic priors from large-scale foundation models (e.g., CLIP/DINO), which frequently fail to exploit the intrinsic modality characteristics essential for high-fidelity fusion. To address these issues, we propose P2Fusion, a prior-guided distillation-based framework that reformulates IVIF via dual intrinsic prompts. Instead of imposing hard-coded penalties, we distill image-intrinsic priors, thermal saliency and spatial quality, into learnable dynamic regulators. Specifically, a Teach-to-Fuse mechanism provides dual-granularity progressive guidance, coupled with a Gated Dynamic Expert Recalibration (GDER) module for decoupled feature refinement. This design enables the network to adaptively mediate modal competition through expert specialization. Extensive experiments demonstrate that P2Fusion achieves state-of-the-art performance across five mainstream datasets. Notably, our framework demonstrates consistent performance advantages in fusion quality, achieving state-of-the-art results in 14 out of 20 key evaluation metrics across 5 benchmarks. Furthermore, it effectively contributes to the robustness of downstream perception, such as +3.2% mAP on MSRS, +0.5% mAP on M3FD and +0.9% mAP on DroneVehicle for object detection. Our code will be available at https://github.com/YiShi99/P2Fusion

Abstract

Infrared-Visible Image Fusion

Explore similar work

P2Fusion: Prompt-based Progressive Infrared-Visible Image Fusion via Dual-Prior Distillation

EAPFusion: Intrinsic Evolving Auxiliary Prior Guidance for Infrared and Visible Image Fusion

Bridging Human Evaluation to Infrared and Visible Image Fusion