Image Quality Assessment using Image Details in Frequency Domain

  • Muhammad Ibrar -ul-haque Department of Electrical Engineering, Sir Syed University of Engineering & Technology, Karachi.
  • Muhammad Tahir Qadri
  • Najeeb Siddiqui

Abstract

This research proposes a RR (Reduced Reference) DIQAM (Detailed Image Quality Assessment Meter) for DCT (Discrete Cosine Transform) based compressed images. DCT technique divides image in sub blocks to achieve image compression.Therefore, it degrades the IQ (Image Quality) by introducing the distortions called blockiness and blurriness in the compressed image.In the telecommunication systems
scenario, the systems available bandwidth is limited. The proposed IQ assessment technique requiresfewer image details parameters called RR parameters at the receiver, rather than the complete reference image. This paper suggests a method for receiving end to estimate the objective quality of the received image in frequency domain.The proposed IQ meter starts by taking the image through edge detection
method, then converting it into frequency domain by Fourier transform and estimating the image details. The image details calculations include the vertical and horizontal ac harmonics as well as all other ac coefficients present in the image. It has been shown in the presented work that using dc coefficients with the other ac coefficients further improves the quality assessment.The calculated strength of coded image
details at receiver is compared with the received RR parameter for the estimation of distortions, blockiness and blurriness. The accuracy of the designed RR DIQAM algorithm is proved by correlating the estimated objective values of the distortions with the LIVE image database2 subjective DMOS values of blockiness and blurriness. The results obtained by the proposed technique are well matched with the LIVE database
values and provide 94-96% correlation.

Published
Oct 1, 2017
How to Cite
-UL-HAQUE, Muhammad Ibrar; QADRI, Muhammad Tahir; SIDDIQUI, Najeeb. Image Quality Assessment using Image Details in Frequency Domain. Mehran University Research Journal of Engineering & Technology, [S.l.], v. 36, n. 4, p. 8, oct. 2017. ISSN 2413-7219. Available at: <http://publications.muet.edu.pk/index.php/muetrj/article/view/23>. Date accessed: 17 dec. 2017.