Signal Processing: Image Communication 76 (2019) 11–21 Contents lists available at ScienceDirect Signal Processing: Image Communication journal homepage: www.elsevier.com/locate/image Blind quality assessment metric and degradation classification for degraded document images ✩ Atena Shahkolaei a,∗ , Azeddine Beghdadi b , Mohamed Cheriet a a Synchromedia Laboratory for Multimedia Communication in Telepresence, École de technologie supérieure, Montreal (QC), Canada H3C 1K3 b Department of Computer Science & Engineering, Paris 13, University, France ARTICLE INFO Keywords: No-reference image quality assessment Degraded document images Physical noises Local phase Degradation classification Support vector machine ABSTRACT Epochal documents suffer from several types of noises that accumulate and evolve over time. This significantly affects their quality and makes their storage and the interpretation of their visual content problematic. Digital preservation seems the most viable and the most promising. Moreover, measuring the amount of degradation and quality assessment of degraded documents is highly desirable for applications such as selecting the proper algorithms for enhancement and analysis of document images, filtering the damaged images, tuning the processing algorithms parameters, document repairing, psychological study, etc. The first contribution of this work is the proposition of an efficient Multi-distortion Document Quality Measure (MDQM) for quality assessment of physically degraded document images. The proposed MDQM metric is based on three sets of spatial and frequency image features. These features are extracted from two layers of text and non-text and mapped to the mean opinion scores (MOS) using the regression function. The second contribution of this work is to estimate the probability of four common document image distortion types, namely, paper translucency, stain, readers annotations and worn holes in the degraded images. In our experiment, the correlations of seven no-reference image quality assessment (NR-IQA) metrics with the MOS values are evaluated on two available datasets. It is shown that the performance of MDQM metric is significantly better than the state-of-the-art NR-IQA metrics. Moreover, the experimental results demonstrate that MDQM metric not only leads to high efficacy for classification of the various degradations but also maintains a remarkable run-time efficiency. It is worth to mention that the proposed method has been conducted for Arabic documents 1. Introduction Historical document images represent an important part of the cul- tural heritage of countries and civilizations. Therefore, preserving and protecting these cultural heritages is of great importance and respon- sibility of governments and international organizations like UNESCO. In recent decades, digitizing historical documents and manuscripts to preserve and make them accessible via electronic media has received a considerable amount of attention [1]. Although the issue of digitizing these documents is mostly solved, the problem of analyzing them is still an ongoing challenge. The image quality of these documents can be assessed subjec- tively and objectively. One factor that may affect the readability and interpretation of such media is image quality. Although subjective image quality assessment (IQA) assessment is the most reliable method, it demands human participants which makes the assessment time- consuming, tedious and expensive. The average result of a set of ✩ No author associated with this paper has disclosed any potential or pertinent conflicts which may be perceived to have impending conflict with this work. For full disclosure statements refer to https://doi.org/10.1016/j.image.2019.04.009. ∗ Corresponding author. E-mail addresses: atena.shahkolaei.1@ens.etsmtl.ca (A. Shahkolaei), beghdadi@univ-paris13.fr (A. Beghdadi), mohamed.cheriet@etsmtl.ca (M. Cheriet). standard subjective tests is called MOS values. Therefore, objective assessment is the primary choice for IQA applications. Objective IQA automates the estimation of image quality by substituting the human perception process with some quality metrics. Objective assessment methods can be classified into three main categories according to the availability of the reference image. These categories are (1) full refer- ence (FR), (2) reduced-reference (RR) and (3) no reference (NR) [2]. For FR metrics both original and distorted images are available [3–5]. RR metrics use the partial information about both the reference and degraded images [6–8]. Finally, for the NR methods, the evaluation of quality is based on some features and properties of the degraded image without referring to the original one. However, very often a priori knowledge of the distortions is used in the design of NR-IQA [9–13]. In recent years, several objective NR-IQA metrics have been pro- posed in the literature for different applications. In the following, we provide a brief review of NR-IQA metrics. https://doi.org/10.1016/j.image.2019.04.009 Received 30 July 2018; Received in revised form 9 April 2019; Accepted 10 April 2019 Available online 24 April 2019 0923-5965/© 2019 Published by Elsevier B.V.