Fig 1.
Overview of the proposed method for image quality evaluation.
(a) The detected keypoints are represented by red dots. (b) Corresponding points are determined based on the features of the keypoints. (c) Calculation process under the proposed metric using keypoints and features.
Fig 2.
Image data set used for the validation.
Target images with (a) different noise characteristics and (b) different resolution characteristics.
Fig 3.
Overview of the image set used for usefulness verification.
Targets are images with (a) different levels of translation, (b) different degrees of rotation angles, and (c) different levels of scales.
Fig 4.
Changes in feature point correspondences with variations in image noise levels.
When the image quality changes, the features no longer match properly, even at the same position in the image. The left and right sides represent the reference and target images, respectively, for each algorithm. The red dots represent the keypoints extracted for each image, and the green lines represent matches between the keypoints.
Table 1.
Summary of the number of extracted keypoints and number of matching points under different algorithms for images with varying dose levels.
Fig 5.
Scatter plots of the SSIM and proposed index obtained in the validation when there were changes in the (a) exposure dose levels and (b) intensity of Gaussian processing. Statistically significant correlations were obtained for all algorithms in both conditions.
Table 2.
Summary of the correlation coefficient under different algorithms for images with varying image quality.
Table 3.
Comparison of calculation times when using each algorithm.
Fig 6.
Scatter plots of the SSIM and proposed index in the usefulness verification.
(a) There are changes in the amount of translation, (b) there are changes in the angle of rotation processing, (c) several downscaling transformations have been performed, and (d) several enlargement processing steps have been performed.
Table 4.
Summary of the correlation coefficient under different algorithms for images with varying image quality.