Prozkoumání evolučních algoritmů pro analýzu očí v obrazech

Abstract

This diploma thesis explores possibilities of evolutionary algorithms in the field of image processing focused on eye analysis. Acquired informations can be used for gaze estimation or psychic state assessment of a person. In this work, eyelid detector based on differential evolution (DE) is proposed. DE is used to find best fitting parabola in edge-detected image. Also pupil and iris detection method using the SOMA algorithm is proposed. This work discusses necessary image preprocessing for both methods as well. Evaluation of both algorithms is done by comparison to manually made ground truth. Outcomes show that evolutionary algorithms can be used for described problems. In some cases, pupil localization can even work with higher precision compared to selected known methods. Unfortunately, both methods report high time complexity.