Signal Compression and its Applications

Introduction

In today’s digital world, signal compression in the domains of speech, image, audio and video is an important factor in designing multimedia applications. One can even say that signal compression is the backbone of day-to-day technological activities. In this blog, we will discuss the important aspects of signal and data compression and subsequently put forward some of its key applications. But before that, let’s take a look at why exactly is signal compression so crucial.

Table of Contents

1. Why use Signal Compression?

2. What is Signal Compression?

3. Signal Compression Techniques

4. Where do we see it?

5. Conclusion

Why use Signal Compression?

While there are many benefits of compressing a signal, we will discuss about 5 of them:

1. Modern technology is evolving, and so should you. Speech, audio, image, and video signal compression (or low bit rate coding) is a critical technology for fast expanding potential in multimedia products and services.

2. For ease of transmission and storage. Most digital signals used for practical purposes carry a large amount of data. For example, a grayscale image (basically a black&white image) of size 512 x 512 with 8 bits per pixel contains (512)2 x 8 = 2097152 bits, while the number of bits in a same-sized colour image is 3 times that of the grayscale one. Thus to reduce the number of bits present in the image, data compression is required.

3. Another point to note is that data in uncompressed form often requires more bandwidth to transmit it over a network compared to compressed data, which in turn increases the incurred expenses.

4. Reduction of bandwidth rapidly leads to degradation of signal thus causing loss of information. This would lead us to believe that digital methods are less efficient than the analog ones.

5. In most signals, some percentage of the data is not relevant and the information can still be extracted from the signal even if this irrelevant data is removed. Signal compression lets us isolate only those bits of data that are required for processing the information

What is signal compression?

To define the compression in a very simple and lucid language, we can refer to it as the process of reduction of the amount of data, while preserving the information (overall meaning of the) content of the signal to a certain fidelity. But, for science students or enthusiasts who are interested in mathematical representation of the concept, it is the representation of N signal values, each of which is quantized to b bits.

There are two types of signal compression:

1. Lossy compression:

As the name suggests, it is the compression which is not required to reproduce the original bit stream without error. Whatever goes in, does not necessarily come out as it is. It is used where some bits are more significant and some are not. They strive to reproduce a similar signal and not the identical. 

Example, speech signals. While talking on a telephone call, we listen to another person. We are always interested in whatever another person is talking about and not the quality of his/her voice. That means, we are interested in the words of the speaker (information) and we are tolerant to losses until and unless we are not losing any of the information. 

2. Lossless Compression: 

Lossless compression is something exactly opposite to lossy compression. It is the compression which is required to reproduce the original bit stream without error. Whatever goes in, has to come out as it is. It is used where all bits are of equal significance. They strive to reproduce the exactly identical signal. 

Example, music. While listening to music, we are not only interested in lyrics only but also the sound of guitar playing at the back, chorus humming at the back, pleasant flute and every other sound associated with that music. That means, we are interested in the information  as well as the rest of the data associated with it. Hence, we have to consider the losses as well. Will you ever love to listen to Badshah without his music? Certainly not.

Signal Compression Techniques

For those thinking signal compression is achieved through some black magic voodoo thing, well we’re sorry to say, you’re wrong. Most digital signals that we encounter daily display some form of data redundancy, which means some bits of data do not contribute any meaningful information to the overall information conveyed by the signal. Image compression aims to nullify this data redundancy, which can be of three types:

1. Coding redundancy - 

The representation of information is linked to coding redundancy.

Codes serve as a representation of the information.

When an image's grey levels are coded in a way that employs more code symbols to represent each grey level than is strictly necessary, the resulting picture is said to have coding redundancy.

2. Inter-pixel spatial redundancy -

The correlation between nearby pixels in a picture causes interpixel redundancy.  This indicates that pixels in close proximity are not statistically independent. The grey levels aren't all equally likely.  Any pixel's value may be predicted based on the value of its neighbours, indicating that they are strongly linked.  Individual pixels carry only a modest amount of information. The difference between neighbouring pixels can be used to represent a picture to eliminate interpixel redundancy.

3. Psycho-visual redundancy -

Because human vision does not include a quantitative study of every pixel or brightness value in an image, psychovisual redundancies arise. It is only feasible to exclude genuine visual information since the information is not required for regular visual processing.

Where do we see it?

Signal compression is used to send the signal through a medium without utilising it to its maximum extent or to store the signal in a device by consuming less space. To understand this, consider the following images; but before that it should be known to us that an image is also a signal with two-dimensions i.e. an image has length and breadth. An entire image is represented by pixels which have colour values (distinct numbers used to represent distinct colour) to make up the desired image. Image compression basically reduces the number of pixels.

Here in the following picture we have two images, original and compressed one.

  

              Original image         Compressed image

Here, we can observe that the clarity of the picture shown in the rectangle is too good in case of uncompressed original signal, whereas we observe a pixelated image in the rectangle when it is compressed. This is lossy compression. As we have discarded some of the pixels, the size and the quality of the image has reduced significantly; but, we can identify that the image is of a petal of a (strange) flower. 

Before compression:

Original Compressed

Here, we observe a reduction in size of the original image to get a compressed image. 703 KB is much lower than the original 2.83MB (2897.92 KB). With such a reduced size, transfer and storage of signal becomes more convenient, efficient and easy.

 

Conclusion

To summarise, compression of signals in any form can drastically reduce the size of data as well as make the usage of multimedia devices faster. We hope you found this blog insightful and would love to read your feedback below in the comments. Happy reading!

Acknowledgement

This work was fully supported by Sardar Patel Institute of Technology. We would like to thank Prof. Kiran Talele, Prof. Najib Ghatte and the faculties teaching Foundation of Signal Processing for their constant support.

Code Link:

https://drive.google.com/drive/folders/1hBd9fpfzqMWftrfplATtmP93o7y5ad9_?usp=sharing

Video Link:

https://youtu.be/K4TIQBgCmSg