AbstractThe boundaries of fish school echo image, which is usually displayed in black and white shades, are not well defined and the necessity for clear edge display has long been felt. Edge detection is also important in the continuous recognition of the sea bed or in the measurement of the school size. Unfortunately, a typical echo sounder image fluctuates severely, prohibiting edge detection with the original image. We propose an edge detection method in which the image is smoothed first by the two‐dimensional exponential smoothing method, and the difference of expected values of data obtained during this processing is compared with a given threshold value to determine the edge. Three ways of taking the difference are conceivable, namely, difference in the radial direction, in the lateral direction, and in the oblique direction; each may be considered as the difference of two‐dimensional data or a weighted slope average. The proposed method effectively performs the conventional smoothing by masking and the edge detection by Laplacian processing simultaneously. Thus, real‐time edge detection of school or sea bed images has become possible in a simple way. This paper describes the theory of edge detection by the above three methods and compares their effectiveness.