Detecção e histórico de nevos em imagens da pele humana

Abstract

This work aims at the use of computer vision, as from digital image processing, for the computational development of detection methods, categorization, obtainment and storage of information of spots present on human skin. The project has goal to develop a prototype of controlled image acquisition, for detecting, quantization and categorization of gifts spots, with purpose in contributing to the researches and prevention of skin cancer Technologies. The development work is scoped to the study to detect spots through the development of image processing algorithms, subsequent to obtaining results that provide information for storage in the image database, aiming to develop a historic of changes arising in the skin. Were developed tests using C ++ algorithms for processing for color and grayscale. The processing was performed by color segmentation between background, skin and spots. The processing by gray-scale tones was made by use of edge identification filter as a basis for processing to identify the spots. The feature vector used was comprised of following data: The number of identified regions, in which for each region was extracted the number of pixels, the average and standard deviation of each color channel and the minimum and maximum coordinates of the axes X and Y. The extracted features of images are stored in a database and the images were stored using the metadata method.