Smoothing

Just like in one-dimensional signals, we are always susceptible to receiving some noise in our images and we generally apply some preprocessing filters to them before we perform our main work on the images. We can consider noise as a random variation of color or brightness information that is not present in the imaged object, which can take place undesirably due to a sensor and circuitry of a digital camera or scanner. This section uses the ideas of low-pass filter kernels to smoothen our images. These filters remove high frequency content, such as edges and noises, although some techniques allow edges not to be blurred. We will cover the four main image filters available in OpenCV: averaging, Gaussian, median filtering, and bilateral filtering.

Note

2D Kernel Convolution is a form of mathematical convolution. An output image is calculated by sweeping each of the pixels of a given image and applying a kernel operator to them, yielding an output pixel for each resulting operation...