Search

Image segmentation is a fundamental process in image processing, aiming to transform the image into a more comprehensible or simplified representation. This process groups pixels under common characteristics, facilitating the unique identification of regions or elements of interest. The quality of segmentation is crucial as it significantly influences the subsequent stages of image processing. Image segmentation plays a vital role in various advanced applications, including computer vision and the analysis of medical, topographical, and astronomical images. Given that there is no universal segmentation method guaranteeing optimal performance for all cases, selecting an appropriate technique for a specific type of image or application represents a complex and demanding challenge. In this work, we propose the use of Multi-Objective Evolutionary Algorithms (MOEAs) as a training tool that integrates operations representative of the usual techniques and strategies in image segmentation. This allows for the generation of operation sequences adapted to specific applications or types of images. The objective functions used to guide the evolutionary process are the maximization of sensitivity (TPR) and specificity (TNR), fundamental components of ROC analysis. Experiments were conducted with multiple images sharing common characteristics from image databases, specifically: i ) benign and malignant melanoma images, ii ) ophthalmoscopic retinal images, and iii ) binary cell form images. We compared the segmentation generated by our proposed algorithm with the ideal segmentation. The results are quite promising and demonstrate the feasibility of using MOEAs to generate sequences of segmentation operations valid for specific applications.

An important aspect of the design of effective machine learning algorithms is the complexity analysis of classification problems. In this paper, we propose a study aimed at determining the relation between the number of adjacent inputs with different labels and the required number of examples for the task of inducing a classification model. To this aim, we first quantified the adjacent inputs with different labels as a property, using a measure denoted as Neighbour Input Variation (NIV). We analyzed the relation that NIV has to random data and overfitting. We then demonstrated that a threshold of NIV may determine if a classification model can generalize to unseen data. We also presented a case study aimed at analyzing threshold neural networks and the required first hidden layer size in function of NIV. Finally, we performed experiments with five popular algorithms analyzing the relation between NIV and the classification error on problems with few dimensions. We conclude that functions whose similar inputs have different outputs with high probability, considerably reduce the generalization capacity of classification algorithms.