Emotion editing method adjusts images to match target feelings precisely
What We Talk About When We Talk About Emotion Editing
Computer Vision and Pattern Recognition
Summary
Changing the emotion an image shows is tricky because the same scene can feel many ways. The authors studied how brain-based models predict what feelings an image gives and used this info to guide image editing. They created a new method that keeps the important parts of the picture while changing its emotion accurately. This method was tested on hundreds of examples and showed it could often make the edited image feel the intended emotion without losing the original content.
What this means in practice
- •For graphic designers: Edit images to express specific emotions while preserving visual content for use in advertising and media.
- •For social media content creators: Automatically adjust the emotional mood of photos to increase engagement or match campaign themes without manual edits.
Authors
Xingzu Zhan, Lin Gu, Ruogu Fang
Abstract
Emotional image editing aims to change the emotion a scene conveys while preserving its essential content. Because the same scene can support different emotional interpretations, achieving an emotional goal requires visual changes grounded in the source context. Motivated by the Component Process Model, which links emotional responses to event appraisal, we investigate neural representations as a reference for these editing decisions. Using TRIBE v2, an fMRI encoding model, we analyze 400 source-target pairs across eight emotion categories. Hidden states from EmoEdit, EmoEditor, and AffectDelta correspond more strongly to predicted target-image responses in visual regions than in affect-related regions. This imbalance motivates ROI-ReFT, which combines source visual responses as a content reference with a predicted affective response change as an editing direction. Two interacting branches use both conditions to guide a frozen editor throughout generation. Predicted responses of the generated image supervise affective change and constrain visual drift. On a post hoc selected subset of 400 editing requests, ROI-ReFT achieves 91.75% target-emotion accuracy and 83.25% joint success, with the same output both matching the target emotion and reaching source-output DINO-I of at least 0.80. Our code and dataset will be publicly released.