Gaussian Blur Algorithm

Gaussian blur is an algorithm for smoothing a digital image by convolving it with a two dimensional Gaussian function, the same bell shaped curve named for the mathematician Carl Friedrich Gauss, so that each output pixel becomes a weighted average of its neighborhood with the nearest pixels weighted most heavily and more distant ones weighted progressively less. Because the two dimensional Gaussian function factors cleanly into the product of two one dimensional Gaussian functions, the blur can be computed as one pass along the image's rows followed by one pass along its columns rather than as a single more expensive two dimensional operation, a property called separability that makes the algorithm efficient even at large blur radii. The technique is used throughout image processing to reduce noise before further analysis, to soften an image for visual effect, and to prevent aliasing artifacts when an image is being reduced in size, and it commonly serves as the preprocessing smoothing step inside other algorithms, including the Canny edge detector.