Outperforming Traditional and Deep Learning Algorithms: A Novel Approach for Background Reconstruction in Intermittent Videos

This study investigates the effectiveness of a novel model for background reconstruction in intermittent videos. Table 3 presents the Average Gray Error (AGE) scores of the proposed model, several advanced traditional algorithms (CVU, MPU, SPMD), and a deep learning algorithm. The algorithms were evaluated on a dataset comprising 16 representative intermittent videos.

The results indicate that the proposed model achieves the best results on five out of the 16 videos, a performance matched by the CVU algorithm. The MPU and SPMD algorithms achieve the best results on three and two videos, respectively. Notably, the deep learning algorithm only achieves the best result on a single video (Teknomo).

These findings suggest that the proposed model demonstrates superior performance compared to some advanced traditional algorithms in handling intermittent video background reconstruction. Furthermore, its performance is comparable to, and in some cases surpasses, that of existing deep learning algorithms. This highlights the progressive nature of the proposed model and its potential for significant contributions to the field.