What does convolution refer to in the context of computer vision?

In the context of computer vision, convolution specifically refers to an operation that combines two functions to create a third function, which highlights specific features in the data. This mathematical process takes two inputs: an image (the first function) and a filter or kernel (the second function), and produces an output known as a feature map. The convolution operation is fundamental in convolutional neural networks (CNNs), which are a prevalent architecture used for image recognition and classification tasks.

By applying the convolution operation, the network is capable of detecting patterns, edges, and textures within an image, enabling the extraction of meaningful features essential for tasks like object detection or segmentation. The ability to capture local spatial hierarchies is a significant advantage of using convolution, as it allows the model to learn progressively more complex features at each layer.

This understanding is crucial for grasping how CNNs function, as they rely heavily on this operation to process visual information effectively.