Measurement of image degradation: a no-reference approach

Article Type

Research Article

Publication Title

Multimedia Tools and Applications

Abstract

In modern communication systems, transmission of data, particularly that of images is very important. Unfortunately, the process is prone to different kinds of degradations. For instance, the large traffic on the network makes the reduction of data volume and storage imperative. This is achieved through compression techniques such as JPEG and JPEG2000, at the expense of image quality, unfortunately. Image quality also deteriorates on account of the state of the transmission channel and the amount of losses occurring. Another common type of image degradation is blurring. At the receiving end, identifying the degradation is an important requirement for applications like blind image denoising so that appropriate techniques for removal may be adopted. Since the undegraded original image is usually not available, no-reference methods are not only attractive but also essential. This article proposes a novel no-reference method which can clearly distinguish between noise, compression and blurring. This is achieved in two stages – the first distinguishes between noise and compression/blurring while the second distinguishes between compression and blurring. Finally, a quantification of the degradation suffered is also carried out, which forms another important contribution of the approach. The primary tool employed is the Discrete Wavelet Transform (DWT) and the differential behaviour of the image DWT coefficients to noise, compression and blurring, at various levels of resolution. The use of the properties of the inter-level coefficients make this approach unique in nature. These properties have been explored and analyzed in detail. A large number of simulations using images from popular databases, establish the usefulness of the approach.

First Page

5545

Last Page

5572

DOI

10.1007/s11042-019-08444-w

Publication Date

3-1-2020

This document is currently not available here.

Share

COinS