Projector Photometric Compensation for Multi-Surface

In this paper proposed for the first time the MS-PCNet network for adaptive photometric compensation of projectors in multiple surfaces. In this paper integrate the multiple attention and triple interactive attention mechanisms in the feature extraction module, using a skip connections multi-level encoder-decoder architecture, which performs matrix addition and subtraction operations on feature maps of the same size from two encoders, and establishes multi-scale interactions between multiple projection surfaces and projected images, so that the network can learn different feature information at different levels to achieve accurate inference and compensation. The experimental results show that the MS-PCNet network exceeds the current state-of-the-art in both single and multiple surfaces, and in this paper method has better compensation. Keywords-Projector compensation; Multiple attention mechanisms; Projector-camera systems; triple interactive attention mechanisms; triple interactive attention mechanisms.