Metodologia baseada em warping para correção de distorções em sistemas de endoscopia

Abstract

Images captured in endoscopy examinations show some distortion, the radial is the most visible. The radial distortion appears in images due to lens used in endoscopes. When a doctor or researcher is analyzing a distorted image, he can not see the exact size of an organ, tumor or lesion, due the magnification in the center of the image and a contraction in the peripheral regions caused by such distortion. This paper aims to propose a new method for correction of radial distortion in endoscopic images. The proposed method is based on concepts of Morphing and Warping, which are techniques widely used in computer graphics to transform the image of one person to another or to cause objects deformations. The system presents the advantage of automatic application of these techniques, since the vast majority of correction algorithms need user interaction. The proposed method uses a pattern image, created for the calibration of the transformations that are applied to the images of endoscopy. The developed system receives the pattern image captured by endoscope, divides the image into a triangular mesh, make the matching of each triangle of the meshes and store the affine transformations of each triangular region of the mesh separately. After calibration, the affine transformations of each triangular region are used in real image of endoscopic examinations performed by the same endoscope used for calibration to correct the images. Finally, the method was compared with others in the literature and has been made quantitative and qualitative analysis with the results.