MeSegHI: um método de segmentação para o processamento linear e não-linear de imagem

Abstract

This research aims to develop the hybrid segmentation s method which uses two techniques: linear mean-shift and non-linear image foresting transform watersheds (IFT). To implement this method the reduced color set was used, it was obtained by mean-shift application, like markers to IFT algorithm. With this technique, the user doesn t need to concern about selecting the correct set marker to Watershed-IFT segmentation s process, since hybrid method will do so automatically. If the seeds are not good enough or are placed just outside the region of interest, the user can add r remove such seeds in order to improve watershed segmentation results. With the purpose of developing a robust and efficient system, with high degree of reusability, it was decided to make use of the object-oriented paradigm together design patterns application. The Java programming language was used to implement the hybrid s method because it supports object oriented and it has an Application Programming Interface (API) Java Advanced Imaging (JAI) that allows an easier image processing operations implementation. The Unified Modeling Language (UML) was used to design the system; it helped the developer to extract the best necessary information to construct the software. After it develops the hybrid segmentation method were realized three applications to prove hybrid method efficience. The first application helps the patologystis to classify Papanicolau examination in yours respectives levels; the second makes a comparison between Watersheds-IFT and hybrid segmentation methods; and the third aim to help profissionals of medicine area to measure the depth of the skin cancer in histologys tissues.