VisionX V4VSFVisionX V4
NAME

vsf − very general spatiotemporal filter

SYNOPSIS

vsf [if=infile] [of=ofile] [k=kfile] [-e] [-d] [-f] [-g]

DESCRIPTION

Vsp convolves a small kernel matrix(image) with an image, the result has the same dimensions as the image. All computations are done in floating point and no scaling is performed on the result. It is the users responsibility to see that no overflow occurs. Only the first image found in each input frame is processed.

When the kernel contains several images a spatiotemporal filtering operation is performed; if the -pd flag is specified the number of image frames output is reduced by (the number of kernel images - 1).

Vsp can also perform the basic morphological operations of erosion and dilation. Zero values in the kernel are taken as non kernel locations for these operations. By default a Binary morphological oeration is performed where zero represents false and anything else represents true. The -g option specifies umbra grey level operation.

By default the kernel matrix is centered on each pixel of the input file.

CONSTRAINTS

Multispectral images are currently not supported. All numeric image types are supported for both kernel and input images.

OPTIONS

k=<file> specifies the kernel file for the convolution.

-c

specifies no automatic centering. Alignment will be according to the bounding box of the kernel and the image; i.e. index 0,0 of the kernel will be aligned with each pixel of the image.

-d

specifies a dilation operation

-e

specifies an erosion operation

-f

This option forces the output to be of type float independent of the input image type. In this case just the image data is output; any other information in the frame is discarded.

-g

Perform grey level morhological operations -pd do not pad at begining and end (z dimesnion)

-w

very verbose, debug mode flag

SEE ALSO

vconv(1)

AUTHOR

A. P. Reeves