Affective Image Quality Evaluation of Smartphone Camera Images

Abstract

Smartphone cameras have become the primary tools for everyday image capture, yet their quality is still mostly optimized using signal-based metrics rather than how users actually see and choose images. This dissertation examines emotion-based image quality evaluation for smartphone camera images and focuses on how perception-based preference arises from affective judgments of images and videos. The work adopts an affective engineering framework in which human observers rate visual stimuli using carefully selected keywords such as natural, realistic, lively and HDR-ness, together with explicit preference scores, and these perceptual responses are related to controllable parameters of the imaging pipeline. This study investigates how users evaluate smartphone image quality in three application domains: selfie photography, chroma-enhanced photographs and high dynamic range video. In the selfie study, front camera images were captured with contemporary smartphones under controlled lighting conditions showed that facial skin color attributes were the dominant determinants of selfie preference. The chroma enhancement study analyzed everyday photographs across several content categories and demonstrated that preferred image quality emerges from a balance between naturalness and lively, and that the optimal chroma level depends on scene category. The third study examines high dynamic range video and introduces the concept of perceived HDR-ness as a specific affective dimension of image quality. Across the two tone-mapping experiments, one using smartphone-generated HDR videos on a tablet display and one using parametric tone curves on an HDR-capable television, resulted in two dominant perceptual dimensions, a shadow-detail dimension and a brightness-liveliness dimension. Preference tended to increase with higher brightness and liveliness. Scene-wise analyses showed that several tone-curve settings produced higher preference than the base HDR image, with the magnitude of improvement depending on scene content. Taken together, the findings reveal that preference for smartphone camera images and videos is not a single judgment, but a perception-based outcome that arises from affective dimensions. Across the three application domains, preference did not arise from a single factor but from a balance among affective attributes, whose relative importance varied with image content. By organizing these dimensions and their relationship to technical parameters of capture, the dissertation provides practical guidance for the design of smartphone camera processing that better reflects user perception and supports more satisfying visual experience under realistic usage conditions.