An Enhanced Differential Evolution Algorithm with Sorted Dual Range Mutation Operator to Solve Key Frame Extraction Problem

This paper proposes a modified Differential Evolution (DE) algorithm in which the conventional mutation operation of DE is replaced by a ‘sorted population’ based mutation operation. This ‘sorted population’ based mutation operation, proposed by authors, differs from the conventional mutation operation in the way in which it selects the candidates for the mutation process and the values it sets for the mutation scale factor (F). The modified DE was implemented, to verify its superiority, on solving 14 different standard benchmarking problems. A comparative study, based on the results obtained, revealed that the proposed algorithm solved the problems providing optimal solutions with lesser time, for higher dimensional problems. Next, the experiments were extended to solve the key frames problem from videos. This part of the experiment combined the conventional SSIM (Structural Similarity Index) approach of key frame extraction with the proposed DE. The results showed that the proposed DE was giving comparatively better results than classical DE.