Is structural awareness the key to event camera data cleansing for enhancing veracity?

Neuromorphic vision sensors, also known as event cameras, offer significant advantages over conventional frame-based cameras, in terms of high dynamic range, low latency, and low power consumption. However, their high sensitivity to illumination changes and asynchronous operation introduces substantial data (typically in the form of false or structure-irrelevant event data) posing challenges to the veracity of the acquired information in downstream vision tasks, such as, object recognition, feature tracking and human action recognition. Traditional cleansing methods for neuromorphic vision sensor event data typically rely on denoising techniques guided by reference-based metrics, which require auxiliary modalities such as Active Pixel Sensor (APS) frames or manual annotations. These references are often unavailable in real-world scenarios. Moreover, existing reference-free metrics generally overlook structural integrity, leading to deceptively high scores when aggressive noise removal results in the loss of meaningful structure. In this paper, we propose the Temporal Structural Event Index (TSEI), a novel, reference-free, structure-aware metric designed to assess the veracity of cleansed neuromorphic vision sensor event data. TSEI integrates temporal structural similarity (TSSM) and contrast normalization within an adaptive segmentation framework to jointly evaluate signal preservation and noise suppression. Experiments on both synthetic and real-world datasets demonstrate that TSEI strongly correlates with structural fidelity and recognition accuracy, outperforming existing metrics in detecting over-cleansing and structural degradation. These findings highlight that structural awareness is a critical factor in enhancing the veracity of neuromorphic event data and ensuring reliable performance in visual recognition tasks.