<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2 Final//EN"> <!--Converted with LaTeX2HTML 98.1p1 release (March 2nd, 1998) originally by Nikos Drakos (nikos@cbl.leeds.ac.uk), CBLU, University of Leeds * revised and updated by: Marcus Hennecke, Ross Moore, Herb Swan * with significant contributions from: Jens Lippmann, Marek Rouchal, Martin Wilck and others --> <HTML> <HEAD> <TITLE>The convolution from the continuous wavelet transform</TITLE> <META NAME="description" CONTENT="The convolution from the continuous wavelet transform"> <META NAME="keywords" CONTENT="vol2"> <META NAME="resource-type" CONTENT="document"> <META NAME="distribution" CONTENT="global"> <META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1"> <LINK REL="STYLESHEET" HREF="vol2.css"> <LINK REL="next" HREF="node330.html"> <LINK REL="previous" HREF="node328.html"> <LINK REL="up" HREF="node328.html"> <LINK REL="next" HREF="node330.html"> </HEAD> <BODY > <!--Navigation Panel--> <A NAME="tex2html5556" HREF="node330.html"> <IMG WIDTH="37" HEIGHT="24" ALIGN="BOTTOM" BORDER="0" ALT="next" SRC="icons.gif/next_motif.gif"></A> <A NAME="tex2html5553" HREF="node328.html"> <IMG WIDTH="26" HEIGHT="24" ALIGN="BOTTOM" BORDER="0" ALT="up" SRC="icons.gif/up_motif.gif"></A> <A NAME="tex2html5547" HREF="node328.html"> <IMG WIDTH="63" HEIGHT="24" ALIGN="BOTTOM" BORDER="0" ALT="previous" SRC="icons.gif/previous_motif.gif"></A> <A NAME="tex2html5555" HREF="node1.html"> <IMG WIDTH="65" HEIGHT="24" ALIGN="BOTTOM" BORDER="0" ALT="contents" SRC="icons.gif/contents_motif.gif"></A> <BR> <B> Next:</B> <A NAME="tex2html5557" HREF="node330.html">The Wiener-like filtering in</A> <B> Up:</B> <A NAME="tex2html5554" HREF="node328.html">Noise reduction from the</A> <B> Previous:</B> <A NAME="tex2html5548" HREF="node328.html">Noise reduction from the</A> <BR> <BR> <!--End of Navigation Panel--> <H2><A NAME="SECTION002061000000000000000"> The convolution from the continuous wavelet transform</A> </H2> We will examine here the computation of a convolution by using the continuous wavelet transform in order to get a framework for linear smoothings. Let us consider the convolution product of two functions: <BR> <DIV ALIGN="CENTER"> <!-- MATH: \begin{eqnarray} h(x)=\int_{-\infty}^{+\infty} f(u)g(x-u)dx \end{eqnarray} --> <TABLE ALIGN="CENTER" CELLPADDING="0" WIDTH="100%"> <TR VALIGN="MIDDLE"><TD NOWRAP ALIGN="RIGHT"><IMG WIDTH="277" HEIGHT="73" ALIGN="MIDDLE" BORDER="0" SRC="img767.gif" ALT="$\displaystyle h(x)=\int_{-\infty}^{+\infty} f(u)g(x-u)dx$"></TD> <TD> </TD> <TD> </TD> <TD WIDTH=10 ALIGN="RIGHT"> (14.68)</TD></TR> </TABLE></DIV> <BR CLEAR="ALL"><P></P> We introduce two real wavelets functions <IMG WIDTH="51" HEIGHT="44" ALIGN="MIDDLE" BORDER="0" SRC="img768.gif" ALT="$\psi(x)$"> and <IMG WIDTH="48" HEIGHT="44" ALIGN="MIDDLE" BORDER="0" SRC="img769.gif" ALT="$\chi(x)$">such that: <BR> <DIV ALIGN="CENTER"> <!-- MATH: \begin{eqnarray} C=\int_0^{+\infty} \frac{\hat{\psi}^*(\nu)\hat{\chi}(\nu)}{\nu}d\nu \end{eqnarray} --> <TABLE ALIGN="CENTER" CELLPADDING="0" WIDTH="100%"> <TR VALIGN="MIDDLE"><TD NOWRAP ALIGN="RIGHT"><IMG WIDTH="232" HEIGHT="83" ALIGN="MIDDLE" BORDER="0" SRC="img770.gif" ALT="$\displaystyle C=\int_0^{+\infty} \frac{\hat{\psi}^*(\nu)\hat{\chi}(\nu)}{\nu}d\nu$"></TD> <TD> </TD> <TD> </TD> <TD WIDTH=10 ALIGN="RIGHT"> (14.69)</TD></TR> </TABLE></DIV> <BR CLEAR="ALL"><P></P> is defined. <I>W</I><SUB><I>g</I></SUB>(<I>a</I>,<I>b</I>) denotes the wavelet transform of <I>g</I> with the wavelet function <IMG WIDTH="50" HEIGHT="44" ALIGN="MIDDLE" BORDER="0" SRC="img771.gif" ALT="$\psi(x)$">: <BR> <DIV ALIGN="CENTER"> <!-- MATH: \begin{eqnarray} W_g(a,b)=\frac{1}{\sqrt a}\int_{-\infty}^{+\infty}g(x)\psi^*(\frac{x-b}{a})dx \end{eqnarray} --> <TABLE ALIGN="CENTER" CELLPADDING="0" WIDTH="100%"> <TR VALIGN="MIDDLE"><TD NOWRAP ALIGN="RIGHT"><IMG WIDTH="365" HEIGHT="73" ALIGN="MIDDLE" BORDER="0" SRC="img772.gif" ALT="$\displaystyle W_g(a,b)=\frac{1}{\sqrt a}\int_{-\infty}^{+\infty}g(x)\psi^*(\frac{x-b}{a})dx$"></TD> <TD> </TD> <TD> </TD> <TD WIDTH=10 ALIGN="RIGHT"> (14.70)</TD></TR> </TABLE></DIV> <BR CLEAR="ALL"><P></P> We restore <I>g</I>(<I>x</I>) with the wavelet function <IMG WIDTH="48" HEIGHT="44" ALIGN="MIDDLE" BORDER="0" SRC="img773.gif" ALT="$\chi(x)$">: <BR> <DIV ALIGN="CENTER"> <!-- MATH: \begin{eqnarray} g(x)=\frac{1}{C}\int_0^{+\infty}\int_{-\infty}^{+\infty}\frac{1}{\sqrt a} W_g(a,b)\chi(\frac{x-b}{a})\frac{dadb}{a^2} \end{eqnarray} --> <TABLE ALIGN="CENTER" CELLPADDING="0" WIDTH="100%"> <TR VALIGN="MIDDLE"><TD NOWRAP ALIGN="RIGHT"><IMG WIDTH="455" HEIGHT="73" ALIGN="MIDDLE" BORDER="0" SRC="img774.gif" ALT="$\displaystyle g(x)=\frac{1}{C}\int_0^{+\infty}\int_{-\infty}^{+\infty}\frac{1}{\sqrt a} W_g(a,b)\chi(\frac{x-b}{a})\frac{dadb}{a^2}$"></TD> <TD> </TD> <TD> </TD> <TD WIDTH=10 ALIGN="RIGHT"> (14.71)</TD></TR> </TABLE></DIV> <BR CLEAR="ALL"><P></P> The convolution product can be written as: <BR> <DIV ALIGN="CENTER"> <!-- MATH: \begin{eqnarray} h(x)=\frac{1}{C}\int_0^{+\infty}\frac{da}{a^{\frac{5}{ 2}}}\int_{-\infty}^{+\infty} W_g(a,b)db\int_{-\infty}^{+\infty}f(u)\chi(\frac{x-u-b}{a}) du \end{eqnarray} --> <TABLE ALIGN="CENTER" CELLPADDING="0" WIDTH="100%"> <TR VALIGN="MIDDLE"><TD NOWRAP ALIGN="RIGHT"><IMG WIDTH="591" HEIGHT="73" ALIGN="MIDDLE" BORDER="0" SRC="img775.gif" ALT="$\displaystyle h(x)=\frac{1}{C}\int_0^{+\infty}\frac{da}{a^{\frac{5}{ 2}}}\int_{-\infty}^{+\infty} W_g(a,b)db\int_{-\infty}^{+\infty}f(u)\chi(\frac{x-u-b}{a}) du$"></TD> <TD> </TD> <TD> </TD> <TD WIDTH=10 ALIGN="RIGHT"> (14.72)</TD></TR> </TABLE></DIV> <BR CLEAR="ALL"><P></P> Let us denote <!-- MATH: $\tilde{\chi}(x)=\chi(-x)$ --> <IMG WIDTH="138" HEIGHT="44" ALIGN="MIDDLE" BORDER="0" SRC="img776.gif" ALT="$\tilde{\chi}(x)=\chi(-x)$">. The wavelet transform <I>W</I><SUB><I>f</I></SUB>(<I>a</I>,<I>b</I>) of <I>f</I>(<I>x</I>) with the wavelet <!-- MATH: $\tilde{\chi}(x)$ --> <IMG WIDTH="48" HEIGHT="44" ALIGN="MIDDLE" BORDER="0" SRC="img777.gif" ALT="$\tilde{\chi}(x)$"> is: <BR> <DIV ALIGN="CENTER"> <!-- MATH: \begin{eqnarray} \tilde W_f(a,b)=\frac{1}{\sqrt a}\int_{-\infty}^{+\infty}f(x)\tilde{\chi}(\frac{x-b}{a})dx \end{eqnarray} --> <TABLE ALIGN="CENTER" CELLPADDING="0" WIDTH="100%"> <TR VALIGN="MIDDLE"><TD NOWRAP ALIGN="RIGHT"><IMG WIDTH="357" HEIGHT="73" ALIGN="MIDDLE" BORDER="0" SRC="img778.gif" ALT="$\displaystyle \tilde W_f(a,b)=\frac{1}{\sqrt a}\int_{-\infty}^{+\infty}f(x)\tilde{\chi}(\frac{x-b}{a})dx$"></TD> <TD> </TD> <TD> </TD> <TD WIDTH=10 ALIGN="RIGHT"> (14.73)</TD></TR> </TABLE></DIV> <BR CLEAR="ALL"><P></P> That leads to: <BR> <DIV ALIGN="CENTER"> <!-- MATH: \begin{eqnarray} h(x)=\frac{1}{C}\int_0^{+\infty}\frac{da}{a^2}\int_{-\infty}^{+\infty} \tilde W_f(a,x-b)W_g(a,b)db \end{eqnarray} --> <TABLE ALIGN="CENTER" CELLPADDING="0" WIDTH="100%"> <TR VALIGN="MIDDLE"><TD NOWRAP ALIGN="RIGHT"><IMG WIDTH="459" HEIGHT="73" ALIGN="MIDDLE" BORDER="0" SRC="img779.gif" ALT="$\displaystyle h(x)=\frac{1}{C}\int_0^{+\infty}\frac{da}{a^2}\int_{-\infty}^{+\infty} \tilde W_f(a,x-b)W_g(a,b)db$"></TD> <TD> </TD> <TD> </TD> <TD WIDTH=10 ALIGN="RIGHT"> (14.74)</TD></TR> </TABLE></DIV> <BR CLEAR="ALL"><P></P> Then we get the final result: <BR> <DIV ALIGN="CENTER"> <!-- MATH: \begin{eqnarray} h(x)={1\over C}\int_0^{+\infty}\tilde W_f(a,x)\otimes W_g(a,x)\frac{da}{ a^2} \end{eqnarray} --> <TABLE ALIGN="CENTER" CELLPADDING="0" WIDTH="100%"> <TR VALIGN="MIDDLE"><TD NOWRAP ALIGN="RIGHT"><IMG WIDTH="375" HEIGHT="73" ALIGN="MIDDLE" BORDER="0" SRC="img780.gif" ALT="$\displaystyle h(x)={1\over C}\int_0^{+\infty}\tilde W_f(a,x)\otimes W_g(a,x)\frac{da}{ a^2}$"></TD> <TD> </TD> <TD> </TD> <TD WIDTH=10 ALIGN="RIGHT"> (14.75)</TD></TR> </TABLE></DIV> <BR CLEAR="ALL"><P></P> In order to compute a convolution with the continuous wavelet transform: <UL> <LI>We compute the wavelet transform <!-- MATH: $\tilde W_f(a,b)$ --> <IMG WIDTH="84" HEIGHT="51" ALIGN="MIDDLE" BORDER="0" SRC="img781.gif" ALT="$\tilde W_f(a,b)$"> of the function <I>f</I>(<I>x</I>) with the wavelet function <!-- MATH: $\tilde{\chi}(x)$ --> <IMG WIDTH="49" HEIGHT="44" ALIGN="MIDDLE" BORDER="0" SRC="img782.gif" ALT="$\tilde{\chi}(x)$">; <LI>We compute the wavelet transform <I>W</I><SUB><I>g</I></SUB>(<I>a</I>,<I>b</I>) of the function <I>g</I>(<I>x</I>) with the wavelet function <IMG WIDTH="50" HEIGHT="44" ALIGN="MIDDLE" BORDER="0" SRC="img783.gif" ALT="$\psi (x)$">; <LI>We sum the convolution product of the wavelet transforms, scale by scale. </UL> <P> The wavelet transform permits us to perform any linear filtering. Its efficiency depends on the number of terms in the wavelet transform associated with <I>g</I>(<I>x</I>) for a given signal <I>f</I>(<I>x</I>). If we have a filter where the number of significant coefficients is small for each scale, the complexity of the algorithm is proportional to <I>N</I>. For a classical convolution, the complexity is also proportional to <I>N</I>, but the number of operations is also proportional to the length of the convolution mask. The main advantage of the present technique lies in the possibility of having a filter with long scale terms without computing the convolution on a large window. If we achieve the convolution with the FFT algorithm, the complexity is of order <IMG WIDTH="90" HEIGHT="41" ALIGN="MIDDLE" BORDER="0" SRC="img785.gif" ALT="$N\log_2N$">. The computing time is longer than the one obtained with the wavelet transform if we concentrate the energy on very few coefficients. <P> <HR> <!--Navigation Panel--> <A NAME="tex2html5556" HREF="node330.html"> <IMG WIDTH="37" HEIGHT="24" ALIGN="BOTTOM" BORDER="0" ALT="next" SRC="icons.gif/next_motif.gif"></A> <A NAME="tex2html5553" HREF="node328.html"> <IMG WIDTH="26" HEIGHT="24" ALIGN="BOTTOM" BORDER="0" ALT="up" SRC="icons.gif/up_motif.gif"></A> <A NAME="tex2html5547" HREF="node328.html"> <IMG WIDTH="63" HEIGHT="24" ALIGN="BOTTOM" BORDER="0" ALT="previous" SRC="icons.gif/previous_motif.gif"></A> <A NAME="tex2html5555" HREF="node1.html"> <IMG WIDTH="65" HEIGHT="24" ALIGN="BOTTOM" BORDER="0" ALT="contents" SRC="icons.gif/contents_motif.gif"></A> <BR> <B> Next:</B> <A NAME="tex2html5557" HREF="node330.html">The Wiener-like filtering in</A> <B> Up:</B> <A NAME="tex2html5554" HREF="node328.html">Noise reduction from the</A> <B> Previous:</B> <A NAME="tex2html5548" HREF="node328.html">Noise reduction from the</A> <!--End of Navigation Panel--> <ADDRESS> <I>Petra Nass</I> <BR><I>1999-06-15</I> </ADDRESS> </BODY> </HTML>