Content-Preserving Principal Feature Blending for Training-Free Style Personalization

Le-Hoang Bui; Hoang-Nguyen Cao; Trung-Nghia Le; Minh-Triet Tran; Dinh-Khoi Vo

cs.CVOct 31, 2026

Content-Preserving Principal Feature Blending for Training-Free Style Personalization

Authors: Le-Hoang Bui, Hoang-Nguyen Cao, Trung-Nghia Le, Minh-Triet Tran, Dinh-Khoi Vo

Organizations: University of Science, Ho Chi Minh City, Vietnam · Vietnam National University, Ho Chi Minh City, Vietnam

Abstract

Reference-based style personalization should reproduce the appearance of a style reference while preserving the subject specified by a content prompt. In visual autoregressive models, dominant reference features can also carry object and spatial information, causing content leakage. We propose Content-Preserving Principal Feature Blending (CoPFB), a training-free method that estimates a low-rank content basis from a clean generation branch and removes the content-aligned component of the style residual. CoPFB applies rank-one blending at the first two autoregressive scales, followed by weaker content-projected refinement at the next three scales. On 1,900 combinations of Parti content prompts and FineStyle reference images, CoPFB improves text alignment from 0.293 to 0.297 and the harmonic score from 0.391 to 0.393 over PFB+SAC, while style similarity decreases from 0.590 to 0.577. CoPFB requires no fine-tuning or learnable parameters. Project page: https://vdkhoi20.github.io/CoPFB/.

Figures & tables

Figure 1: CoPFB results. Each row shows a style reference followed by outputs for different content prompts.

Panel of Figure 2 — Figure 2: Coarse-to-fine dynamics of next-scale prediction.

Panel of Figure 3 — Figure 3: Quantitative and qualitative identification of the pivotal scale.

Figure 5: CoPFB overview. Black arrows show the clean content path; purple arrows show generation and blending.

Variant	CP	SAC	$S_{\mathrm{txt}}\uparrow$	$S_{\mathrm{img}}\uparrow$	$S_{\mathrm{harmonic}}\uparrow$
PFB only	–	–	0.255	0.572	0.353
PFB+SAC	–	$\checkmark$	0.293	0.590	0.391
PFB+CP	$\checkmark$	–	0.284	0.563	0.378
PFB+CP+SAC	$\checkmark$	$\checkmark$	0.295	0.561	0.387

Table 3: Component ablation. All variants use the same PFB backbone.

$r_{c}$	$S_{\mathrm{txt}}\uparrow$	$S_{\mathrm{img}}\uparrow$	$S_{\mathrm{harmonic}}\uparrow$
1	0.295	0.561	0.387
2	0.288	0.554	0.379
4	0.281	0.547	0.371

Table 4: Effect of content-basis rank across rank-1, rank-2 and rank-4.

Image Style Transfer

Image style transfer changes an image’s visual appearance to match an artistic, photographic, or other reference while preserving its content.

See all 63 papers

CLeaR: A Unified Framework for Resolving the Leakage-Degradation Dilemma in Style Transfer
Sep 29, 2026Teng Zhou, Yunhao ChenImage Style Transfer
Zhejiang University · Fudan University
FreeStyle: Free Control of Style-Content Dual-Reference Generation from Community LoRA Mining
Jun 18, 2026Jinghong Lan, Wei Cheng, Yunuo Chen +10Image Style Transfer Image Generation
Fudan University · StepFun · University of Hong Kong +1
StyleGallery: Training-free and Semantic-aware Personalized Style Transfer from Arbitrary Image References
Mar 11, 2026Boyu He, Yunfan Ye, Chang Liu +3Image Style Transfer Diffusion-Based Image Editing
College of Computer Science and Technology, National University of Defense Technology · School of Design, Hunan University

Sep 29, 2026J/K move · Enter open · S save

CLeaR: A Unified Framework for Resolving the Leakage-Degradation Dilemma in Style Transfer

Teng Zhou, Yunhao Chen

Zhejiang University · Fudan University

Image Style Transfer

Open paper AskPDF arXiv

Style transfer aims to render target content in the style of a reference image, but existing methods often suffer from content leakage, where objects, layouts, or semantics from the style reference appear in the generated output. Although prior data-driven and training-free methods can reduce leakage, they often face a leakage-degradation dilemma: stronger content suppression may weaken style fidelity, while richer style preservation may reintroduce unwanted reference content. We identify this dilemma across the full style-transfer pipeline, including feature separation, feature-space grounding, and diffusion generation. To address these issues, we propose CLeaR, a training-free framework for content-leakage-resistant style transfer. CLeaR first uses Orthogonal Subspace Projection to define content-reduced style targets in each vision foundation model (VFM) feature space. It then performs Ensemble Inversion, which optimizes a shared pixel-space style anchor satisfying style constraints across multiple VFMs. Finally, Energy-Guided Calibration maintains style alignment during diffusion sampling by steering the denoising trajectory toward the ensemble-defined style manifold. We further provide a theoretical analysis showing that the style-anchor estimation error decreases with the number of VFMs. Experiments on StyleBench demonstrate that CLeaR improves style alignment, reduces content leakage, and achieves better LLM-as-Judge evaluation compared with existing methods. The code is available at https://github.com/0606zt/CLeaR.

Abstract

Image Style Transfer

Explore similar work

CLeaR: A Unified Framework for Resolving the Leakage-Degradation Dilemma in Style Transfer

FreeStyle: Free Control of Style-Content Dual-Reference Generation from Community LoRA Mining

StyleGallery: Training-free and Semantic-aware Personalized Style Transfer from Arbitrary Image References