Sistema PIIC: Procesamiento de imágenes con Inteligencia Computacional
In this paper, the design and implementation of a system called PiiC is presented. This system allows image segmentation using predicate logic with fuzzy truth values, according to preset conditions and after being trained by an expert user. The main objective is making intuitive the application of fuzzy predicates for a non-expert user, achieving “translating" the natural language to an algorithm that uses a fuzzy mathematical model. The system allows defining one or more predicates to describe regions or objects in the image through the image characteristics (features), determined pixel by pixel, and their fuzzy terms associated with them. The same number of predicates as regions or objects to be segmented is required. Quantitative measures of quality of the obtained segmentation are also obtained using a segmented image by experts. The knowledge stated on predicates can also be reused to process similar images. As future work, automatic discovery of predicates by defining the areas to be segmented is proposed, by using the features and their associated fuzzy terms.