Which process is used to convert a grayscale image into a binary image, and…

Concept

A binary image has only two pixel values (foreground and background). The standard way to obtain one from a grayscale image is thresholding — every pixel above a chosen intensity threshold becomes one value and every pixel below it becomes the other. Separately, the Canny edge detector is defined by a three-stage pipeline: compute the intensity gradient, apply non-maximum suppression to thin the edges to one-pixel width, and apply hysteresis (double) thresholding to link strong edges and reject weak isolated ones.

Application

Applying this to the question: the process that turns a grayscale image into a binary image is thresholding, and the edge detector built from gradient computation, non-maximum suppression, and hysteresis thresholding is the Canny algorithm. So the matching pair is Thresholding, Canny.

Contrast with the distractors

• Quantization reduces the number of intensity levels represented but does not by itself produce a strictly two-level (binary) image; Sobel estimates only the gradient, with no non-maximum-suppression or hysteresis stage.

• Sampling sets the spatial resolution (pixel grid) of an image rather than reducing its values to two; Prewitt, like Sobel, is a plain gradient operator without non-maximum suppression or hysteresis thresholding.

• Compression reduces the storage size of the image data rather than binarizing pixel values; Roberts is a simple 2x2 gradient cross-operator without non-maximum suppression or hysteresis thresholding.

• Laplacian is a second-derivative (zero-crossing) operator that responds to curvature change in intensity, not the gradient-magnitude, non-maximum-suppression, and hysteresis pipeline described in the question.