In the paper quality metric of a test image is designed using fuzzy regression analysis by modeling membership functions of interval type 2 fuzzy set representing quality class labels of the image. The output of fuzzy regression equation is fuzzy number from which crisp outputs are obtained using residual error defined as the difference between observed and estimated output of the image. In order to remove human bias in assigning quality class labels to the training images, crisp outputs of fuzzy numbers are combined using weighted average method. Weights are obtained by exploring the nonlinear relationship between the mean opinion score (MOS) of the image and defuzzified output. The resultant metric has been compared with the existing quality metrics producing satisfactory result.
CITATION STYLE
De, I., & Sil, J. (2015). A fuzzy regression analysis based no reference image quality metric. Advances in Intelligent Systems and Computing, 320, 87–95. https://doi.org/10.1007/978-3-319-11218-3_9
Mendeley helps you to discover research relevant for your work.