Electroencephalography (EEG) is highly susceptible to ocular artefacts from eye movements and blinks, which contaminate frontal channels and degrade brain-computer interface (BCI) performance. Artefacts are commonly mitigated using electrooculog- raphy (EOG) as an auxiliary channel, but extra periocular electrodes increase setup time and are often absent in recordings, whereas many BCI systems already capture face video. This research asks whether such video can provide an EOG-like surrogate for ocular artefact handling in EEG. VEOS (Vision-based Electrooculogram Signal Inference) is a framework that maps monocular eye-region video to an inferred EOG signal. A theoretical model links eyeball and eyelid kinematics, their two-dimensional image landmarks, and the one- dimensional electrical signals observed in EOG and frontal EEG. A multimodal plat- form was implemented to record synchronised face video, EEG, and horizontal and vertical EOG during controlled ocular tasks, forming the VEOS development dataset. An established open dataset providing high-density EEG, EOG, and monocular eye video, collected independently by other researchers, serves as an independent test bed for evaluating generalisation. VEOS uses dense facial landmarks to estimate eye-centred pupil position and eyelid aperture. Pupil and canthus locations are recovered with errors below one eye width, while noisier eyelid landmarks still support blink timing and aperture estimation. Temporal neural networks operating on eye pose and aperture achieve correlations of 0.75–0.8 with EOG, outperforming linear baselines. The inferred EOG is integrated into a regression-based linear component analysis pipeline. VEOS recovers roughly 60–70% of the blink suppression obtained with true EOG in frontal EEG, preserves motor imagery, steady-state visual evoked potential and P300 features within a few percent, and yields BCI accuracy gains comparable to real EOG. The research shows that single-camera, vision-based EOG inference is feasible and useful for EEG artefact mitigation, while not fully replacing dedicated EOG channels.