Fast Image Quality Assessment via Hash Code
No-reference image quality assessment (NR-IQA) is significant for image processing and yet very challenging, especially for real-time application and big image data processing. Traditional NR-IQA metrics usually train complex models such as support vector
- PDF / 392,428 Bytes
- 10 Pages / 439.37 x 666.142 pts Page_size
- 43 Downloads / 218 Views
Abstract. No-reference image quality assessment (NR-IQA) is significant for image processing and yet very challenging, especially for realtime application and big image data processing. Traditional NR-IQA metrics usually train complex models such as support vector machine, neural network, and probability graph, which result in long training and testing time and lack robustness. Hence, this paper proposed a novel no-reference image quality via hash code (NRHC). First, the image is divided into some overlapped patches and the features of blind/ referenceless image spatial quality evaluator (BRISQUE) are extracted for each patch. Then the features are encoded to produce binary hash codes via an improved iterative quantization (IITQ) method. Finally, comparing the hash codes of the test image with those of the original undistorted images, the final image quality can be obtained. Thorough experiments on standard databases, e.g. LIVE II, show that the proposed NRHC obtains promising performance for NR-IQA. And it has high computational efficiency and robustness for different databases and different distortions. Keywords: No-reference
1
· Image quality assessment · Hash code
Introduction
With the tremendous development of intelligent network, ultra-high resolution display, and wearable devices, high quality and credible visual information (image, video, etc.) is significant for the end user to obtain a satisfactory quality of experience (QoE). Where, assessing the quality of visual information, especially no-reference or blind image quality assessment (NR-IQA or BIQA), plays an important role in numerous visual information processing system and applications [1]. Moreover, effective (high quality prediction accuracy) and efficient (low computational complexity) NR-IQA is essential and has attracted a large number of attentions. NR-IQA metric is designed to automatically and accurately predict image quality without reference images. Hence, it is a difficult and challenging work and has attracted many researchers’ attentions. Traditional methods focus on designing distortion-specific methods [2]-[4], which means that these methods evaluate images with only one kind of distortions effectively, such as JPEG compression, JPEG2000 compression, white noise, and Gaussian blurring. Therefore, c Springer-Verlag Berlin Heidelberg 2015 H. Zha et al. (Eds.): CCCV 2015, Part II, CCIS 547, pp. 276–285, 2015. DOI: 10.1007/978-3-662-48570-5 27
Fast Image Quality Assessment via Hash Code
277
it is imperative to build the general purpose NR-IQA metric to handle different types of distortions and even multi-distortions. Recently, great effort has been made to design general purpose NR-IQA metrics. A series of methods are presented in the literature [5]-[18]. Almost all of the reported NR-IQA methods include quality-aware feature extraction and effective evaluation model designing, which are the most important processing for building a NR-IQA method. Generally, natural scene statistical (NSS) properties [19] are most popular utilized features, which ar
Data Loading...
