Motion-Induced Perceptual Blur in Stabilized Videos: Modeling, Quantification, and Application

In low-light video capture, long exposure times increase intra-frame image blur, and hold-type displays introduce motion-dependent display blur during playback. Video stabilization reduces inter-frame motion and thus suppresses display blur, decreasing the overall blur compared with unstabilized videos. However, stabilized low-light videos can paradoxically suffer from frame-to-frame fluctuations of perceptual blur. This effect cannot be explained solely by image or display blur. Instead, it arises from the interaction between the two types of blur, producing visually objectionable blur artifacts in stabilized videos. This perceptual interaction effect is referred to as Motion-Induced Perceptual Blur (MIPB). This study introduces an MIPB metric, a perceptually grounded measure that explicitly quantifies the interaction between image blur and display blur. To demonstrate practical utility, an MIPB-aware motion injection method is developed for stabilized videos, which suppresses MIPB by constraining its value below a user-defined threshold while maintaining acceptable stability. Through extensive experiments on both synthetic and real-world videos, the proposed metric is confirmed to reliably capture MIPB characteristics across diverse lighting and stability conditions. The motion injection method is established to effectively mitigate MIPB while preserving stability. Subjective MOS evaluation further verifies that the proposed MIPB metric aligns well with human perception.