We are comparing the features of two images of the same person using two different cameras, and would like to determine the level of similarity between the two images. We have already used the Bland-Altman plot, which confirms that the two are not dissimilar. Linear-regression shows that there is a close relationship. However, we are not sure if this translates to 'being similar'.
Tests such as Anova is not appropriate as there are large inter-subject variability.
What would be a suitable method to test if the features of the images from the two cameras are similar?
Comparing features in two pictures may not be sufficient... 1)Take multiple pictures (of the same objects under identical circumstances per camera pair) 2) select the most pertinent features for what relates to identity or difference... 3) assess the measures for meeting the assumptions of the inferential approach 4) do the analysis of variance (related samples, for the approach that I suggested)
After you conduct steps 1 and 2 as suggested by Simon, you can run t-test to test for equality of means and F-test to test for equality of variances.
Of course, if you want to take the extra step and make use of more advanced methodologies, consider starting with digital files from the images, and then convert each file with wavelet functions. Then, using the new data sets that are created with wavelet functions, you will have over 100,000 data values per image. Then the fun will really start!
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA