Structural Coding

Multimedia computational communications in the context of 6G have the potential to significantly improve transmission efficiency by incorporating intelligent computation into the communication process. This intelligent communication architecture encompasses multimedia representation, coding, transmission, and other aspects from a semantic perspective. The core component of this architecture is multimedia semantic representation, which primarily aims to reduce the amount of multimedia data. In this chapter, we propose the use of sketch graphs as an effective representation of images, capturing pixel variations, geometric feature distribution, and structural information. Sketch graphs hold promise for applications in multimedia computational communications. Specifically, we develop a learning-based method that utilizes deep neural network (DNNs) to extract sketch graphs through edge detection, sketch point detection, and sketch line detection. Moreover, we design an end-to-end extraction method that achieves real-time processing. Experimental results on several datasets demonstrate the superior performance of our approach in terms of classification and generation tasks.