Abstract
Holographic-type communication (HTC) is emerging as an important application scenario for the next-generation communication systems. In particular, to provide immersive user experiences, HTC requires to implement 3D visual reconstruction of a physical object by collecting multi-view images from multiple viewpoints. However, it is a dilemma to balance the efficiency and the visual quality of this distributed HTC scenario. To tackle this problem, a distributed multi-view image semantic communication (DMISC) scheme is proposed in this letter. First, a framework consisting of multiple individual semantic encoders and a joint multi-view semantic decoder is proposed to achieve high-quality image recovery by leveraging the correlations among multi-views. Second, a convolutional neural network-based DMISC model is designed to achieve high communication efficiency and visual quality for 3D reconstruction, by combining joint source-channel coding with joint multi-view decoding. Finally, the simulation results are presented to show that our proposed DMISC scheme can achieve better performance than those existing schemes.