J = VL_IMSMOOTH(I,SIGMA) convolves the image I by an isotropic Gaussian kernel of standard deviation SIGMA. I must be an array of doubles. IF the array is three dimensional, the third dimension is assumed to span different channels (e.g. R,G,B). In this case, each channel is convolved independently.
VL_IMSMOOTH() accepts the following options:
Selects between GAUSSIAN and TRIANGULAR kernels. The triangular kernel support has 2*SIGMA-1 sampels. Kernels have unit mass.
Selects between ZERO or CONTINUITY padding method to handle the image boundaries. ZERO extends the input image with zeroes around the border, and CONTINUITY extends the image with constant pixels.
Sets the subsampling step. A subsampling step of STEP pixels causes J(1:STEPS:end, 1:STEPS:end, :) to be computed. This is useful to downsample the image.
See also: VL_HELP().