No abstract

In this paper we propose a new method to deal with the problem of automatic human skin segmentation in RGB color space model. The problem is modeled as a minimum cost graph cut problem on a graph whose vertices represent the image color characteristics. Skin and non-skin elements are assigned by evaluating label costs of vertices associated to the weight edges of the graph. A novel approach based on an energy function defined in terms of a database of skin and non-skin tones is used to define the costs of the edges of the graph. Finally, the graph cut problem is solved in Graphics Processing Units (GPU) using the Compute Unified Device Architecture (CUDA) technology yielding very promising skin segmentation results for standard resolution video sequences. Our method was evaluated under several conditions, indicating when correct or incorrect results are generated. The overall experiments have shown that this automatic method is simple, efficient, and yields very reliable results.

O câncer de mama é o tipo mais comum de câncer no mundo. Todo ano são detectados milhares de casos de câncer, e destes, 25% são de mama. Sabendo que o diagnóstico precoce é crítico para o prognóstico do paciente, novas tecnologias à base de análise de imagens são desenvolvidas para guiar um diagnóstico eficaz e menos invasivo. Nesse artigo, é desenvolvido um novo método de segmentação de imagens das mamas em imagens termográficas utilizando limiarização com refinamento adaptativo. Esse método se mostrou eficaz com aproximadamente 96% de acurácia e 98% de sensibilidade. Além disso, a abordagem proposta é simples de ser implementada computacionalmente, é eficiente e apropriada para aplicações em tempo real.

Breast cancer is the second most common type of cancer in the world. It is estimated that 29.7% of new cases diagnosed in Brazil occur in any structures of the breasts. However, the disease has a good prognosis if detected early. Thus, the development of new technologies to help doctors to provide an accurate diagnosis is indispensable. The goal of this work is to develop a new method to automate parts of computer-aided diagnosis systems, performing the unsupervised segmentation of the Region of Interest (ROI) of infrared breast images acquired in lateral view. The segmentation proposed in this paper consists of three stages. The first stage pre-processes the infrared images of the lateral region of breasts. Later, features are extracted from a descriptor based on Histogram of Oriented Gradients (HOG). Concluding, a Machine Learning algorithm is used to perform the segmentation of the sample. The current method obtained an average of 89.9% accuracy and 94.3% specificity in our experiments, which is promising compared to other works.

Supernovas são explosões super luminosas que caracterizam o fim da vida de uma estrela supermassiva. Redes Neurais Artificiais são ferramentas promissoras para a automação da análise dos dados de supernovas provenientes de telescópios, porém, apresentam uma desvantagem, que é a grande demanda por recursos de hardware necessários para sua implementação. Neste trabalho, propõe-se uma Rede Neural Convolucional otimizada para Detecção de Supernovas em imagens, utilizando a técnica de poda dos parâmetros. O algoritmo obteve uma acurácia de 0,964 e a poda foi capaz de reduzir o tamanho do modelo em até 70%, sem perda significativa de acurácia.