Fusão de imagens multifocais e realce de nitidez: uma contribuição ao diagnóstico automático da tuberculose

Abstract

Tuberculosis is a disease caused by Mycobacterium tuberculosis that remains one of the deadliest diseases in the world, being necessary efforts to combat it. The conventional light bacilloscopy is a sputum smear microscopic examination to diagnose tuberculosis. Automated applications in microscopic images, such as automated bacilloscopy, face a big problem: blurriness in the images acquired. Thus, it is emphasized the importance of using multi-focus image fusion methods to overcome this problem. In this context, this work aims to propose a method to multi-focus image fusion applied to sputum smear microscopy images in order to obtain images where the greatest number of bacilli is in focus and application of sharpness enhancement to the resulting images. This factor has great importance to enabling automated sputum smear microscopy. To find out existing schemes, we performed simulations and analyzes in some of the techniques in the literature. When it was analized the performance of these techniques through the quality metrics Multichanel Edge Information Preservation and Variance, it was possible to propose four new methods based on the best techniques in the literature, among which, the method with the best performance was choose to investigate configurations that would guarantee the best quality of processed. Has also been proposed the use of an unsharp masking to the post-processing of the new methods, and is shown through examples the effectiveness of this process in increased sharpness and visual appearance. As future work, it is suggested to evaluate this method using grayscale images to verify the impact on the processing time and the quality of the fused images, also the evaluation of the applicability of the image registration.