An Intelligent Method for Edge Detection based on Nonlinear Diffusion

Abstract

Edge detection is an important task in the field of image processing with broad applications in image and vision analysis. In this paper, we present a new intelligent computational mechanism using nonlinear diffusion equations for edge detection. Experimental results show that the proposed method outperforms standard edge detectors as well as other methods that deploy inhibition of texture.

Segmentation of Blood Vessels in Retinal Images Based on Nonlinear Filtering

“…We propose to filter the retinal image by an anisotropic diffusion approach [2] which smooths the image while preserves edges intensities related to the vessels. For that purpose, we adjust the filter parameters to decrease the diffusion on the boundaries and to perform a few iterations for blurring homogeneous regions.…”

“…Aiming to obtain smooth vessels boundaries in the segmentation, instead of using the original image, we applied a nonlinear anisotropic diffusion filtering [2], which blurs homogeneous regions while preserves vessels intensities. The retinal background is approximated by a median filtering in the original image.…”

A computer-assisted approach to supporting taxonomical classification of freshwater green microalga images

“…From a wide diversity of possible image filtering techniques, those based on anisotropic diffusion [74] have been successfully applied to images from several domains for noise supression, such as medicine [75] [76], Synthetic aperture radar (SAR) [77] and botany [78]. Perona and Malik [67] formulated a non-linear diffusion method for avoiding the blurring and localization problems of the aforementioned filtering techniques.…”

“…(2.1) and (2.2) are set according to the image characteristics and the desired degree of smoothness. Further details about ADF discretization and implementation and about parameter settings can be found elsewhere [74] [81]. Figure 10 illustrates this filtering process applied to a green algae image, shown in Figure 10(a).…”

“…Generally, automated systems for identifying biological entities employ Gaussian filtering. However, to achieve image smoothing while preserving important boundary information of the target alga regions, the anisotropic diffusion filter was successfully applied [74].…”