Image Quality Assessment Using the Joint Spatial/Spatial-Frequency Representation
- PDF / 1,774,264 Bytes
- 8 Pages / 600.03 x 792 pts Page_size
- 73 Downloads / 277 Views
Image Quality Assessment Using the Joint Spatial/Spatial-Frequency Representation ˘ Azeddine Beghdadi1 and Razvan Iordache2 1 L2TI-Institute 2 GE
Galil´ee, Universit´e Paris 13, 93430 Villetaneuse, France Healthcare Technologies, 78530 Buc, France
Received 9 December 2004; Revised 20 December 2005; Accepted 9 March 2006 Recommended for Publication by Gonzalo Arce This paper demonstrates the usefulness of spatial/spatial-frequency representations in image quality assessment by introducing a new image dissimilarity measure based on 2D Wigner-Ville distribution (WVD). The properties of 2D WVD are shortly reviewed, and the important issue of choosing the analytic image is emphasized. The WVD-based measure is shown to be correlated with subjective human evaluation, which is the premise towards an image quality assessor developed on this principle. Copyright © 2006 Hindawi Publishing Corporation. All rights reserved.
1.
INTRODUCTION
Wigner-Ville distribution (WVD) has been proved to be a powerful tool for analyzing the time-frequency characteristics of nonstationary signals [1]. It is well established that WVD-based signal analysis methods overcome the shortcomings of the traditional Fourier-based methods and that it achieves high resolution in both domains. While WVD is widely used in applications involving 1D signals, the extension to multidimensional signals, in particular to 2D images has not reached a similar development [2]. The use of WVD for image processing was first suggested by Jacobson and Wechsler [3]. It was shown that WVD is a very efficient tool for capturing the essential nonstationary image structures [4, 5]. The interesting properties of joint spatial/spatial-frequency representations of images led to other applications of WVD to image processing, in particular in image segmentation [6–10], demonstrating that WVD-based methods provide high discriminating power for signal representation. Indeed, WVD extracts the intrinsic local spectral features of an image. On the basis of this knowledge, the motivation behind the idea of using WVD for image quality measure is that the extraction and evaluation of a distortion in a given image could be expressed as a segmentation problem. This paper proposes the application of the WVD in analyzing and tracking image distortions for computing an image quality measure. The properties of the 2D WVD and some implementation aspects are briefly discussed.
With the increasing use of digital video compression and transmission systems, image quality assessment has become a crucial issue. In the last decade, there have been proposed numerous methods for image distortion evaluation inspired from the findings on human visual system (HVS) mechanisms [11]. In the vision research community, it is generally acknowledged that the early visual processing stages involve the creation of a joint spatial/spatial-frequency representation [12]. This motivates the use of the WVD as a tool for analyzing the effects induced by applying a distortion to a given image. Depending on the required in
Data Loading...
