The field of Digital Compression Methods for Digital data is becoming more and more important as upcoming applications put tremendous pressures on the storage and transmission capacities of the current technology. Whereas technological advancements in hardware & transmission medium are enhancing our ability to fully utilize the potentials of digital data, the advancements in compression technology enable us to efficiently utilize the capacity of existing hardware.
Digital Compression can be broadly catagorized into two classes: Loss-Less compression and Lossy compression. Loss-less compression methods are based on source coding of data and guarantee that the data is not corrupted or distorted by the compression process. Loss-less compression is a must for binary data in the form of text files, database files, executables etc. Common examples of Loss-Less compression schemes are ZIP, GZIP, ARJ.
Lossy Compression, as the term suggests, distort the data in the compression process but yeild higher compression ratios than their loss-less counterparts. Such methods are used for coding digitized analog signals like voice, audio, images & video. Incidently, these digitized signals are extremely demanding in terms of storage medium and transmission Bandwidth. For these signals, a slight distortion can be tolerated and in certain cases may even be imperceptable to the human auditory/visual system. The key point is that allowing the compression process to distort the signal, we can achieve higher compression ratios and thus lower the storage/transmission requirements to a much larger degree than the capability of loss-less systems. The quality loss associated with lossy coding is directly affected by the compression ratio of the algorithm, with higher compression ratios leading to poorer results.
This project demonstrates a lossy compression system based on the Wavelet Transform. We have used the Discrete Wavelet Transform for Audio Compression & Image compression. Our compression algorithm belongs to the class of Transform based compression schemes where the original signal is transformed into another domain using any linear transform (in our case, the Discrete Wavelet Transform). The transformation step serves to decorrelate the signal and allows us to remove certain portions of the transformed signal that are least important in terms of their effect on output quality. The receiver reconstructs the original signal for the 'truncated' transform by applying the Inverse Transform.
Due to the finite length of Wavelets, the Discrete Wavelet Transform is more suited for compression purposes than the Fourier Transform.
In this page, we outline our work. We will not go into the introductory theory of Wavelets. Several articles & web pages are referred to for this purpose.
Our project is divided into two main portions. Please click on the portion of interest to you for further detail.