Feature extraction by shape matching - Feature Extraction and Image Processing - page 198

if(m>-1 & m<1)
for x0=1:columns
y0=round(ym+m*(x0-xm));
if(y0>0 & y0<rows)
acc(y0,x0)=acc(y0,x0)+1;
end
end
else
for y0=1:rows
x0=round(xm+(y0-ym)/m).
if(x0>0 & x0<columns)
acc(y0,x0)=acc(y0,x0)+1
end
end
end %if abs
end %if M
end

x2=x2=incx;y2=y2+incy;

if x2>x1+I
x2=x1+i;
incx=0; incy=1;
y2=y2+incy;
end

if y2>y1+i
y2=y1+i;
incx=-1; incy=0;
x2=x2+incx;
end

if x2<x1-i
x2=x1-i;
incx=0 incy=-1;
y2=y2+incy;
end
end %for k
end %for I
end %if(x1,y1)
end %y1
end %x1

Code 5.9 Implementation of the parameter space reduction for the Hough transform for ellipses

Figure 5.20 shows the accumulators obtained by the implementation of Code 5.9 for the
images in Figure 5.14 (a) and Figure 5.14 (b). The peak represents the location of the
ellipses. In general, there is noise and the accumulator is wide. This is for two main
reasons. First, when the gradient direction is not accurate, then the line of votes does not
pass exactly over the centre of the ellipse. This forces the peak to become wider with less
height. Second, in order to avoid numerical instabilities we need to select points that are