cv2Canny

cv2.Canny is a function in OpenCV's Python interface used to detect edges in an image by applying the Canny edge detection algorithm. It takes a grayscale image as input and outputs a binary image where edge pixels are highlighted, typically with a value of 255.

The Canny algorithm involves several steps. First, the image is smoothed with a Gaussian filter to reduce noise. Then gradients are computed, usually with the Sobel operator, to estimate edge strength and direction. Non-maximum suppression thins the edge candidates to one-pixel width. Next, double thresholding classifies pixels as strong, weak, or non-edges, and edge tracking by hysteresis preserves true edges while discarding stray responses. In cv2.Canny the gradient computation uses the specified aperture size for the Sobel operator, and a boolean flag selects whether to use the L1 or L2 norm for gradient magnitude.

Input and output: the function expects a single-channel grayscale image. If the input is color, it should be converted to grayscale beforehand. The output is an 8-bit single-channel image where edge locations are set to 255 and non-edges to 0.

Parameters: threshold1 and threshold2 define the low and high thresholds for hysteresis. The apertureSize parameter selects the Sobel kernel size and can be 3, 5, or 7. The L2gradient flag, when true, uses the Euclidean (L2) norm for gradient magnitude; otherwise, the L1 norm is used.

Practical notes: choosing thresholds is image-dependent; common practice is to adjust them interactively or via automatic heuristics. The result can be thick or broken depending on parameters and noise, and post-processing steps like dilation or thinning may be applied as needed. Example usage: edges = cv2.Canny(gray, 50, 150, apertureSize=3, L2gradient=False).