Morphological Auto Adaptive Neuro–Fuzzy System for Automatic Image Segmentation and Edge Detection

Morphological auto Adaptive Neuro–Fuzzy segmentation and edge detection System is presented. This system consists of morphological techniques that perform erosion on an image pixels and performs the image reconstruction operation which reduces the intensity variations of a region and a multilayer perceptron (MLP)-like network that performs image segmentation by adaptive thresholding of the input image using thresholds automatically pre-selected by fuzzy C-means clustering algorithm (FCM), architecture is feed forward, learning is unsupervised. The output status of the network is described as a fuzzy set. Fuzzy entropy is used as measure of the error of the segmentation system as well as a criterion for determining potential edge pixels. From the edge image of the proposed system find the the connected components by convolving the image with a mask, by giving the values between 0 (zero) and max value of image, we can extract all connected components of the image by storing it in an array which are the segments of an image with similar intensity values in each segment. Such an algorithm is most useful for applications that are supposed to work with different types of images. This system is also useful for applications dealing with more complex scenes, where several objects have to be detected.