Smartphone portrait quality improves with detailed face contrast metrics
Interpretable Face Contrast Modeling for Smartphone Portrait Rendering
Computer Vision and Pattern Recognition
Summary
Smartphone pictures of people’s faces often look different depending on the phone used because each phone processes the image uniquely. The authors created a way to break down and measure face contrast into three parts: the overall shading on the face, how eyes stand out, and the fine details in the skin. They tested these measures with people’s opinions and found their method matches human preferences better than older ways of measuring contrast. They also made a special dataset using a realistic mannequin photographed by many smartphones to study this in detail. Their work helps in understanding and improving how phones make people’s faces look in portrait photos.
What this means in practice
- •For smartphone camera engineers: Improve facial rendering algorithms by optimizing contrast components to better match user visual preferences across different lighting conditions.$Commercial implications: Enables camera makers to enhance portrait image quality more consistently, differentiating products with superior facial appearance tuning.
- •For photo editing software developers: Incorporate interpretable contrast metrics into retouching tools to adjust specific facial regions for more natural and preferred portrait enhancements.
Authors
Yujin Cho, Benoît Pochon, Jean-Michel Morel, Gabriele Facciolo
Abstract
The quality of smartphone portrait photography has become increasingly important to both manufacturers and users. Recent devices use complex computational pipelines to improve facial rendering under varied lighting. These pipelines vary widely across manufacturers, so the same portrait can look noticeably different across smartphones. These differences affect multiple aspects of facial appearance, making perceptual preference difficult to characterize with a single measure. To address this, we propose an interpretable framework that decomposes face contrast into three components: overall facial tonal separation, eye-region contrast, and micro-contrast of fine skin details. We formulate dedicated metrics for each component and validate them via pairwise human judgments, demonstrating that our approach correlates more closely with human contrast perception than traditional contrast measures. To support this analysis, we introduce a realistic mannequin-based dataset designed for systematic facial contrast evaluation, comprising 508 full-resolution captures from 29 smartphones under diverse lighting, with region-specific masks and extensive pairwise perceptual annotations. As a downstream analysis, we show that combining the three component-wise metrics provides a meaningful basis for modeling balanced facial contrast preference in a controlled setting, and we present a preliminary evaluation on a smaller real-portrait dataset indicating that insights derived from the mannequin study can extend to real images. Our work provides a controlled first step toward defining and evaluating face contrast in portrait imaging.