| 内容提要: |
The main purpose of this paper is to present a
metrology-based view of the image quality assessment (IQA) field.
Three main topics are developed. First, the state of the art in the
field of IQA is presented, providing a classification of some of the
most important objective and subjective IQA methods. Then, some
aspects of the field are analyzed from a metrological point of view,
also through a comparison with the software quality measurement
area. In particular, a statistical approach to the evaluation of the
uncertainty for IQA objective methods is presented, and the topic
of measurement modeling for subjective IQA methods is analyzed.
Finally, a vector approach for full-reference IQA is discussed,
with applications to images corrupted by impulse and Gaussian
noise. For these applications, the vector root mean squared error
(VRMSE) and fuzzy VRMSE are introduced. These vector parameters
provide a possible way to overcome the main limitations of
the mean-squared-error-based IQA methods. |