Advances in Visual Data Compression and Communication

i

i

i

i

i

i

i

i

426 19 Computational Information Theory

19.3 Source Coding

In this section, we first discuss how to efficiently compress the metadata. After that,

we discuss the coding of Cloud image sources and Cloud video sources, respectively.

One more challenging problem is how to compress an input source by using Cloud

sources at the receiver. This is distributed cloud-based coding.

19.3.1 Coding of Metadata

In Shannon’s information theory, the coding of metadata is studied relatively less

than other areas, probably because metadata is easy to compress. However, in com-

putational information theory, there is rich metadata that has to be generated offline

and stored with Cloud sources, including local feature descriptors, source correlation,

scene descriptors, ...