next up previous [pdf]

Next: Coding multidimensional convolution and Up: FILTERING ON A HELIX Previous: Multidimensional deconvolution breakthrough

Examples of simple 2-D recursive filters

Let us associate $ x$ - and $ y$ -derivatives with a finite-difference stencil or template. (For simplicity, take $ \Delta x=\Delta y=1$ .)

$\displaystyle \frac{\partial }{ \partial x } \eq \begin{array}{\vert c\vert c\vert} \hline 1 & -1 \\ \hline \end{array}$ (3)

$\displaystyle \frac{\partial }{ \partial y } \eq \begin{array}{\vert r\vert} \hline 1 \\ \hline -1 \\ \hline \end{array}$ (4)

Convolving a data plane with the stencil (3) forms the $ x$ -derivative of the plane. Convolving a data plane with the stencil (4) forms the $ y$ -derivative of the plane. On the other hand, deconvolving with (3) integrates data along the $ x$ -axis for each $ y$ . Likewise, deconvolving with (4) integrates data along the $ y$ -axis for each $ x$ . Next, we look at a fully 2-dimensional operator (like the cross derivative $ \partial_{xy}$ ).

A nontrivial 2-dimensional convolution stencil is:

\begin{displaymath}\begin{array}{\vert r\vert r\vert} \hline 0 & -1/4 \\ \hline 1 & -1/4 \\ \hline -1/4 & -1/4 \\ \hline \end{array}\end{displaymath} (5)

We convolve and deconvolve a data plane with this operator. Although everything is shown on a plane, the actual computations are done in one dimension with equations (1) and (2). Let us manufacture the simple data plane shown on the left in Figure 3. Beginning with a zero-valued plane, we add in a copy of the filter (5) near the top of the frame. Nearby, add another copy with opposite polarity. Finally, add some impulses near the bottom boundary. The second frame in Figure 3 is the result of deconvolution by the filter (5) using the 1-dimensional equation (2). Notice that deconvolution turns the filter into an impulse, while it turns the impulses into comet-like images. The use of a helix is evident by the comet images wrapping around the vertical axis.

wrap-four
wrap-four
Figure 3.
Illustration of 2-D deconvolution. Left is the input. Right is after deconvolution with the filter (5) as preformed by by module polydiv
[pdf] [png] [scons]

The filtering in Figure 3 ran along a helix from left to right. Figure 4 shows a second filtering running from right to left. Filtering in the reverse direction is the adjoint. After deconvolving both ways, we have accomplished a symmetrical smoothing. The final frame undoes the smoothing to bring us exactly back to where we started. The smoothing was done with two passes of deconvolution, and it is undone by two passes of convolution. No errors, and no evidence remains at any of the boundaries where we have wrapped and truncated.

back-four
back-four
Figure 4.
Recursive filtering backward (leftward on the space axis) is done by the adjoint of 2-D deconvolution. Here we see that 2-D deconvolution compounded with its adjoint is exactly inverted by 2-D convolution and its adjoint.
[pdf] [png] [scons]

Chapter [*] explains the important practical role to be played by a multidimensional operator for which we know the exact inverse. Other than multidimensional Fourier transformation, transforms based on polynomial multiplication and division on a helix are the only known easily invertible linear operators.

In seismology we often have occasion to steer summation along beams. Such an impulse response is shown in Figure 5.

dip
dip
Figure 5.
Useful for directional smoothing is a simulated dipping seismic arrival, made by combining a simple low-order 2-D filter with its adjoint.
[pdf] [png] [scons]

Of special interest are filters that destroy plane waves. The inverse of such a filter creates plane waves. Such filters are like wave equations. A filter that creates two plane waves is illustrated in figure 6.

wrap-waves
wrap-waves
Figure 6.
A simple low-order 2-D filter with inverse containing plane waves of two different dips. One is spatially aliased.
[pdf] [png] [scons]


next up previous [pdf]

Next: Coding multidimensional convolution and Up: FILTERING ON A HELIX Previous: Multidimensional deconvolution breakthrough

2015-03-25