Display omitted •Full-field Experimental Modal Identification on high-speed videos.•LSCF eigenvalue stabilization from accelerometer data and LSFD modal constant from high-speed video.•High frequency (close to 10kHz) mode shapes.•Identification below the noise level. Vibration measurements using optical full-field systems based on high-speed footage are typically heavily burdened by noise, as the displacement amplitudes of the vibrating structures are often very small (in the range of micrometers, depending on the structure). The modal information is troublesome to measure as the structure’s response is close to, or below, the noise level of the camera-based measurement system. This paper demonstrates modal parameter identification for such noisy measurements. It is shown that by using the Least-Squares Complex-Frequency method combined with the Least-Squares Frequency-Domain method, identification at high-frequencies is still possible. By additionally incorporating a more precise sensor to identify the eigenvalues, a hybrid accelerometer/high-speed camera mode shape identification is possible even below the noise floor. An accelerometer measurement is used to identify the eigenvalues, while the camera measurement is used to produce the full-field mode shapes close to 10kHz. The identified modal parameters improve the quality of the measured modal data and serve as a reduced model of the structure’s dynamics.