Optimización de la Segmentación Local de Sauvola Aplicada a la Detección de Defectos Superficiales en Escenas con Iluminación No Homogénea

The presence of non-homogeneous illumination in real scenes images is an actual problem that difficult the correct segmentation of these. This paper presents a methodology for optimizing Sauvola local segmentation for the detection of superficial defects in non-homogeneous illuminated images by adjusting its parameters through genetic algorithms. The methodology consists of these stages: First, the problem is proposed from the perspective of genetic algorithms where each individual in the population represents the values for Sauvola's parameters. Then several fitness functions are proposed using comparison metrics between a Sauvola's segmentation and one performed manually. Each function is evaluated by running the genetic algorithm with it in a subset of images. The best fitness function, according to the results of optimization, is used again in a larger sample. Finally, the last optimization results are analyzed by a clustering analysis. The results show that it is possible to adjust Sauvola's parameters to successfully segment each image but these do not exhibit a tendency to a specific point that allow to suggest unique parameters to segment all images with a high performance.