Sobel Operator Algorithm

The Sobel operator is a discrete differentiation algorithm used in image processing to approximate the gradient of an image's intensity at every pixel, producing an output image in which edges, the points of sharpest brightness change, stand out clearly. It works by convolving the image with two small three by three kernels, one tuned to detect vertical edges by approximating the horizontal gradient and the other tuned to detect horizontal edges by approximating the vertical gradient, each kernel combining a simple differencing step with a triangular smoothing step that makes the result less sensitive to noise than a plain difference would be; the two results are then typically combined to give the overall gradient magnitude and direction at each point. Irwin Sobel and Gary Feldman, both then at the Stanford Artificial Intelligence Laboratory, presented the operator in an internal talk in 1968, and it became one of the most widely taught and implemented edge detection building blocks in image processing.