Localized filters modify the value of each image pixel based on the value of pixels in its neighborhood. Localized filters are sometimes implemented as a convolution. Convolution filtering is a method for modifying the appearance of an image by convolving its pixel values with a transformation kernel. When the kernel is small it is more efficient to use the built-in operation MatrixConvolve. When the convolution kernel is relatively large it is more efficient to compute the convolution using the fast Fourier transform (FFT). A number of typical convolution kernels are implemented in the operation MatrixFilter which also contains a few localized (neighborhood) operators.