Image Restoration under Semantic Communications

TL;DR

This paper introduces a two-stage image reconstruction method in semantic communications that improves image quality by decoding and enhancing images from noisy data, outperforming traditional methods.

Contribution

It proposes a novel two-stage reconstruction process leveraging channel knowledge and image statistics for superior image restoration in semantic communication systems.

Findings

The two-stage process significantly improves image quality over traditional decoding.

Different evaluation metrics may conflict when assessing system performance.

Numerical results confirm the effectiveness of the proposed method.

Abstract

Semantic communication has emerged as the breakthrough beyond the Shannon theorem by transmitting and receiving semantic information instead of data bits or symbols regardless of its content. This paper proposes a two-stage reconstruction process to boost the system's performance. In the first phase, the image information is first decoded from the noisy received data by exploiting the channel knowledge. The decoded image is enhanced by a post-filter and image statistics. Different metrics are exploited to evaluate the image restoration quality of our considered model. Numerical results are obtained using natural images that verify the superior improvements of the proposed two-stage reconstruction process over the traditional decoded data. Moreover, the different metrics assessing the system performance based on their criteria can be conflicted with each other.