00000caa a22000004a 4500 UP-8027390931312542890 Buklod 20210617090354.0 o--- | || || ta 210617s xx d r |||| u| (iLib)UPCEB-00012051722 emz Three-step Approach to Edge Detection of Texts. pp. 193-211 We proposed a three-step image segmentation approach to determine the edges of images containing old texts. In general, texts from old books and articles tend to be very noisy. Thus, we first employed a suitable denoising method to obtain a smooth approximation I_s of a given image I ̃. Then, the fuzzy edge map E ̃ was obtained using the gradient of I_s. This gradient map gave an estimate of the edges of the texts. For the second step, the method of k-means++ with two clusters was employed to separate the edges from rest of the image. Because a smooth approximation of the image was used, the edges obtained are "thick." And so, in the last step of the our method, the binary image generated from the previous step was post-processed using a thinning algorithm. We implemented our method to images containing Baybayin texts from the National Museum of the Philippines. Science Periodicals. Edge detection. Image denoising. Method of k-means ++. Thinning algorithm. Recio, K. R. Philippine Journal of Science. vol. 148, 1 (Mar) 2019. Analytics