Blur-Specific No-Reference Image Quality Assesment for Microscopic Hyperspectral Image Focus Quantification

Abstract

Hyperspectral (HS) imaging is a novel technique that allows for better understanding of materials, being an improvement in multiple applications. However, one of its main drawbacks is the focus assessment. This issue has already been covered for RGB images. Thus, in this study, it is going to be revised several no reference RGB image quality assessment algorithms (NR-IQA). To this aim, a HS image database was created by capturing different images of the same specimen at different working distances. NR-IQA algorithms were tested over monochromatic images extracted from the HS images. Additionally, a study through each independent wavelength was carried out. Results showed that some algorithms perform better for calibration samples and another ones for biological samples. Furthermore, focus differences were found in the initial and final wavelengths. In conclusion, HS image results are similar to the one obtained for RGB images but, there is still room for improvement.