.TH BLIND-APPLY-KERNEL 1 blind
.SH NAME
blind-apply-kernel - Apply a convolution matrix to a video
.SH SYNOPSIS
.B blind-apply-kernel
[-apPxy]
.I kernel-stream
.SH DESCRIPTION
.B blind-apply-kernel
reads a video from stdin and a convolution matrix video
from
.I kernel-stream
and apply the convolution matrix in each frame the
same frame in stdin, and prints the resulting video
to stdout.
.SH OPTIONS
.TP
.B -a
Used to optimise performance if it is known that
the video is opaque, and to ensure that the output
video is opaque.
.TP
.B -p
Each frame in
.I kernel-stream
shall contain one matrix per pixel in a frame in
stdin. The width of
.I kernel-stream
shall be a multiple of the width of stdin, the width
shall be the width of stdin multiplied by the width
of the convolution matrix. The height of
.I kernel-stream
shall be a multiple of the height of stdin, the
height shall be the height of stdin multiplied by the
height of the convolution matrix.
.TP
.B -P
Apply the convolution with first premultiplying
the alpha channel.
.TP
.B -x
When encountering the left or right edge of the video,
wrap around to the opposite edge.
.TP
.B -y
When encountering the upper or lower edge of the video,
wrap around to the opposite edge.
.SH REQUIREMENTS
.B blind-apply-kernel
requires enough free memory to load two full frames
from stdin and one full frame from
.I kernel-stream
into memory. However, if
.I -p
is used, the height of
.I kernel-stream
divide by the height of stdin number of rows from
.I kernel-stream
rather than a full frame from
.I kernel-stream
is loaded into memory.
A frame or row requires 32 bytes per pixel it contains.
.SH SEE ALSO
.BR blind (7),
.BR blind-make-kernel (1),
.BR blind-gauss-blur (1)
.SH AUTHORS
Mattias Andrée
.RI < maandree@kth.se >