Compression artifacts reduction with multiscale tensor regularization
- PDF / 1,468,801 Bytes
- 11 Pages / 439.37 x 666.142 pts Page_size
- 72 Downloads / 235 Views
Compression artifacts reduction with multiscale tensor regularization V. B. Surya Prasath1,2,3,4
· Dang N. H. Thanh5 · Le Minh Hieu6 · Le Thi Thanh7
Received: 19 January 2020 / Revised: 12 October 2020 / Accepted: 23 October 2020 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract We study a multiscale tensor regularization based JPEG decompression artifact removal in digital images. Structure tensor eigenvalues based robust edge map is used within a variable exponent regularization. Variational constrained minimization which combines data fidelity driven by color subsampling and discrete cosine transformation operator is utilized. Experimental results across different compression levels and with various error metrics indicate our proposed method obtains high quality results on cartoon/clip-art and LIVE1 natural image databases. Keywords JPEG deblocking · Artifact reduction · Edge preserving · Multiscale · Structure tensor · Regularization
1 Introduction Joint Photographic Experts Group (JPEG) is a lossy compression widely used in digital image transmission and storage despite the introduction of better compression standard JPEG2000 (Wallace 1992; Rabbani 2002). Blocking artifacts are a common drawback in JPEG compressed images, and especially at low bit rates the artifacts manifest as Gibbs
B
V. B. Surya Prasath [email protected]
1
Division of Biomedical Informatics, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH 45229, USA
2
Department of Pediatrics, University of Cincinnati, Cincinnati, OH, USA
3
Department of Biomedical Informatics, College of Medicine, University of Cincinnati, Cincinnati, OH 45267, USA
4
Department of Electrical Engineering and Computer Science, University of Cincinnati, Cincinnati, OH 45221, USA
5
Department of Information Technology, School of Business Information Technology, University of Economics, Ho Chi Minh City, Vietnam
6
Department of Economics, University of Economics - The University of Danang, Da Nang, Vietnam
7
Department of Basic Sciences, Ho Chi Minh City University of Transport, Ho Chi Minh City, Vietnam
123
Multidimensional Systems and Signal Processing
oscillations near edges due to the cancellation of discrete cosine transform (DCT). Due to this manifestation, compression methods can suffer from artifacts (Kulalvaimozhi et al. 2019). Over the years, various post processing techniques have been proposed to remove the blocking artifacts and they can be grouped into three major categories (Shen and Kuo 1998): (a) adaptive filtering which consider JPEG compression artifacts as noise (Yim and Ik Cho 2005; Xu et al. 2011; Pang et al. 2015; Zhang et al. 2014), (b) regularization methods which use prior smoothness constraints (Alter et al. 2005; Bredies and Holler 2012b), and (c) learning methods which approximate the compressed to uncompressed mapping (Dong et al. 2015), with learning methods performing better with added computational complexity compared to filters and regularization models. Regularization methods
Data Loading...
