< Back to previous page

Publication

Towards Content Independent No-reference Image Quality Assessment Using Deep Learning

Book Contribution - Book Chapter Conference Contribution

The study of image quality assessment (IQA) is divided on natural scene and document images which are processed using different models and quality metrics. This casts challenges for the development of content-independent no-reference (NR) IQA models which can operate on different types of images without requiring information regarding the content of the images. In this paper we propose a unified no-reference image quality assessment (UIQA) model using a deep learning approach, where a generalization of NR IQA across natural scene and document images is achieved using a deep convolutional neural network (DCNN). Without having to discriminate the type of the images, the proposed model can assess the quality of natural scene and document images in a blind and uniform manner. Testing results on two benchmarking datasets demonstrate that the proposed model achieves promising performances competitive with the state-of-the-art simultaneously on natural scene and document images.
Book: 2019 IEEE 4th International Conference on Image, Vision and Computing (ICIVC)
Edition: July-2019
Series: Proceedings of 2019 IEEE 4th International Conference on Image, Vision and Computing (ICIVC)
Volume: July-2019
Pages: 276-280
Number of pages: 5
ISBN:978-1-7281-2326-4
Keywords:no-reference image quality assessment, perceptual score, OCR accuracy, deep convolutional neural network, transfer learning
Accessibility:Open