Automatic Assessment of Image Quality
Towards the Internet of Services: The THESEUS Research Program, Springer, Berlin, (2014)DOI: 10.1007/978-3-319-06755-1_19
Abstract
Huge amounts of film and image content are in archives stored on analog media, slowly fading away. In professional environments new content is captured electronically. There has always been the wish to capture or reformat, to store and transmit images electronically in very high quality, but only today have the sensors been developed in a way that the capturing can be done at reasonable cost for everybody. These high quality levels for still and moving pictures come with high amounts of data. Therefore encoding is necessary for storage and transmission. In general, the encoding process is not lossless, but the aim is to preserve the perceived quality. Until recently, the only way to assess the quality of encoding schemes used to be visual tests, but meanwhile measurement schemes have been standardized in ITU-R and ITU-T which can estimate the perceived quality if some conditions are met: the measurement scheme must have access to the raw, non-coded, digital content, and the encoding scheme must be in a family of well-known coding schemes. Non-reference (NR) measurement and measurement of innovative coding schemes are an open issue. Within THESEUS, two new NR predictors were developed and their performance was evaluated. In this paper, the procedure for and the results of the assessment of a measurement scheme which estimates quality without access to the reference are presented.
