Comparing Modern Compression Techniques & DNN Compression

Introduction

Compression, in all of its forms, allows for more efficient delivery of digital information. There are two main types of compression: Lossless compression, and Lossy compression. As of all things, each situation requires a different type of compression, and within each type, there are multiple ways to approach the task. Modern compression techniques, such as HEVC (High-efficiency video coding, also known as H.265) are lossy compression techniques. This survey will analyze and evaluate the multiple available modern standards of video compression, and disclose the tradeoffs and limitations of each of the formats studied.

Motivation

Compression is needed to support the many multimedia applications used by engineers across diverse fields in ECE. Understanding the tradeoffs between cutting-edge compression techniques is important if one intends to implement new designs under the tight engineering constraints of the industry.

Taxonomy

For each category of compression:

Take 3 examples, find fundamental data from research papers, or if needed clone the repo and get benchmarks ourselves
With gathered data, create visualizations that can compare the findings in an understandable way
Compare the examples using the tradeoff criteria
Make a conclusion about which compression methodology is good to use in which situation
Compare new techniques with a baseline that is as standard as possible (H.264 for video)

Tradeoffs for comparison:

Usability - How easy is it to use? How widespread is the support for the compression?
Compression density - How much does it reduce file size?
Compression Computational efficiency - How computationally demanding is the process? How long does it take to compress?
Support moving forward - How maintained is the methodology?
Lossy vs Lossless