Image segmentation is a key tool in computer vision (i.e., helping computers “see” the meaning in pictures). The article “Efficient Quantum Inspired Meta-Heuristics for Multi-Level True Colour Image Thresholding” (Applied Soft Computing 2017: 472–513) reported a study to compare 10 image segmentation algorithms—six conventional, four inspired by quantum computing. Each algorithm was applied to 10 different images, from an elephant to Mono Lake to the Mona Lisa; the images serve as blocks in this study. Kapur’s method, an entropy measure for image segmentation tools, was applied to each (algorithm, image) pair; lower numbers are better. The article reports the following rank averages for the 10 algorithms.

Does the data indicate that the 10 algorithms are not equally effective at minimizing Kapur’s entropy measure? Test at the .01 significance level. What do the rank averages suggest about quantum-inspired versus conventional image segmentation methods?

GA 8.30 QIGAMLTCI 2.15 DE 6.60 QIDEMLTCI 2.35 SA 9.10 QISAMLTCI 3.65 BSA 5.90 PSO 8.90 QIPSOMLTCI 1.85 CODE 6.20