A novel 2D-ABC adaptive filter algorithm: A comparative study

Kockanat S., KARABOĞA N.

DIGITAL SIGNAL PROCESSING, vol.40, pp.140-153, 2015 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 40
  • Publication Date: 2015
  • Doi Number: 10.1016/j.dsp.2015.02.010
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.140-153
  • Erciyes University Affiliated: Yes


Recently, two dimensional (2D) adaptive filter, which can self-adjust the filter coefficients by using an optimization algorithm driven by an error function, has attracted much attention by researchers and practitioners, because 2D adaptive filtering can be employed in many image processing applications, such as image denoising, enhancement and deconvolution. In this paper, a novel 2D artificial bee colony (2DABC) adaptive filter algorithm was firstly proposed and to the best of our knowledge, there is no study describing 2D adaptive filter algorithm based on metaheuristic algorithms in the literature. At the first stage, in order to analyze the performance and computational efficiency of the novel 2D-ABC adaptive filter algorithm, it was used in the 2D adaptive noise cancellation (ANC) as recommend in literature. For a fair comparison, the competitor 2D adaptive filter algorithms were applied to the same 2D-ANC setup under same condition, such as same Gaussian noise, same filter order or same test images. The results of the novel 2D-ABC adaptive filter algorithm were compared with those of the 2D affine projection algorithms (APA), 2D normalized least mean square (NLMS) and 2D least mean square (LMS) adaptive filter algorithms. At the second stage, to demonstrate the robustness of the novel 2D-ABC adaptive filter algorithm, it was implemented for speckle noise filtering on noisy clinical ultrasound images. The results show that the novel 2D-ABC adaptive filter algorithm has a better performance than the other classical adaptive filter algorithms and its denoising efficiency is quite well on noisy images with different characteristics. (C) 2015 Elsevier Inc. All rights reserved.