Available at: https://digitalcommons.calpoly.edu/theses/659
Date of Award
MS in Electrical Engineering
Contrast enhancement techniques are used widely to improve the visual quality of images. The difference in luminance reflected from two adjacent surfaces creates contrast between the surfaces in the image. The greater the contrast, the easier it is to recognize and differentiate objects in an image. Thus object contrast is an important factor in the perception of the visual quality of an image and in its usefulness for object recognition and image analysis applications.
The focus of this thesis is on studying how well different contrast enhancement techniques developed for visible spectrum photographic images work on infrared images; to determine which techniques might be best suited for incorporation into commercial infrared imaging applications. The database for this thesis consists of night vision infrared images taken using a Photon 320 camera by FLIR Systems, Inc.
Numerous contrast enhancement techniques found in literature were reviewed in this project, out of which four (4) representative techniques have been selected and presented in detail. Homomorphic filtering, fuzzy-logic enhancement, and single-scale retinex techniques have been implemented based on published papers. These were each compared to the classical technique of histogram equalization using metrics of computational time, histogram standard deviation, sharpness and user observations. These metrics provide both quantitative and qualitative analyses of the implemented techniques which are relevant to the end user applications of infrared imaging.
Based on the metric calculations and results, homomorphic filtering and histogram equalization proved to have better results compared to the other techniques. All the implemented techniques are global contrast enhancement methods and for future work local contrast enhancement techniques may be applied to the results obtained in this investigation as a post-processing technique.