Abstract

The present study examines image sharpening techniques quantitatively. A technique known as unsharp masking has been the preferred image sharpening technique for imaging professionals for many years. More recently, another professional-level sharpening solution has been introduced, namely, the high-pass filter technique of image sharpening. An extensive review of the literature revealed no purely quantitative studies that compared these techniques. The present research compares unsharp masking (USM) and high-pass filter (HPF) sharpening using an image quality metric known as Visual Information Fidelity (VIF). Prior researchers have used VIF data in research aimed at improving the USM sharpening technique. The present study aims to add to this branch of the literature through the comparison of the USM and the HPF sharpening techniques. The objective of the present research is to determine which sharpening technique, USM or HPF, yields the highest VIF scores for two categories of images, macro images and architectural images. Each set of images was further analyzed to compare the VIF scores of subjects with high and low severity depth of field defects. Finally, the researcher proposed rules for choosing USM and HPF parameters that resulted in optimal VIF scores. For each category, the researcher captured 24 images (12 with high severity defects and 12 with low severity defects). Each image was sharpened using an iterative process of choosing USM and HPF sharpening parameters, applying sharpening filters with the chosen parameters, and assessing the resulting images using the VIF metric. The process was repeated until the VIF scores could no longer be improved. The highest USM and HPF VIF scores for each image were compared using a paired t-test for statistical significance. The t-test results demonstrated that: • The USM VIF scores for macro images (M = 1.86, SD = 0.59) outperformed those for HPF (M = 1.34, SD = 0.18), a statistically significant mean increase of 0.52, t = 5.57 (23), p = 0.0000115. Similar results were obtained for both the high severity and low severity subsets of macro images. • The USM VIF scores for architectural images (M = 1.40, SD = 0.24) outperformed those for HPF (M = 1.26, SD = 0.15), a statistically significant mean increase of 0.14, t = 5.21 (23), p = 0.0000276. Similar results were obtained for both the high severity and low severity subsets of architectural images. The researcher found that the optimal sharpening parameters for USM and HPF depend on the content of the image. The optimal choice of parameters for USM depends on whether the most important features are edges or objects. Specific rules for choosing USM parameters were developed for each class of images. HPF is simpler in the fact that it only uses one parameter, Radius. Specific rules for choosing the HPF Radius were also developed for each class of images. Based on these results, the researcher concluded that USM outperformed HPF in sharpening macro and architectural images. The superior performance of USM could be due to the fact that it provides more parameters for users to control the sharpening process than HPF.

Publication Date

9-2021

Document Type

Thesis

Student Type

Graduate

Degree Name

Print Media (MS)

Department, Program, or Center

School of Media Sciences (CET)

Advisor

Bruce Leigh Myers

Advisor/Committee Member

Robert Eller

Campus

RIT – Main Campus

Share

COinS