A novel approach to image edge detection using Kalman filtering
Shubhankar Borse, Prabin Kumar Bora · 2016
The estimation theory, in general, was developed to target three major applications. Smoothing, filtering and prediction. Linear estimators, thus can predict future outcomes of an event or set of events, based on a weighted sum of the past set of events. It is obvious that when the outcome has a steep change in the next step, the error in prediction increases. Using this property and applying it to the spatial domain of an image, we know that the edges of the image depict steep changes, and the error of prediction at edges is highest. This paper proposes a method to detect the edges of an image using the estimation theory, and compares it with the existing methods, such as Canny, Sobel and Prewitt.