improc ass4: Finished Gaussian Derivatives and Canny Edge Detector sections in report.

improc/ass4/canny.py

@@ -14,10 +14,10 @@ def canny(F, s, Tl=None, Th=None):
     image F. Optionally specify a low and high threshold (Tl and Th) for
     hysteresis thresholding."""
     # Noise reduction by a Gaussian filter
-    F = gD(F, s, 0, 0)
+    F = gD(F, 1.4, 0, 0)
 
     # Find intensity gradient
-    W = Gauss1(1.4, 1)
+    W = Gauss1(s, 1)
     Gx = convolve1d(F, W, axis=1, mode='nearest')
     Gy = convolve1d(F, W, axis=0, mode='nearest')
     G = zeros(F.shape)
@@ -47,11 +47,12 @@ def canny(F, s, Tl=None, Th=None):
                     and (not in_image(nb, G) or g > G[nb]):
                 E[y, x] = g
 
-    # Only execute hysteresis thresholding if the thresholds are specified
+    # Only execute hysteresis thresholding if both thresholds are specified
     if Tl is None or Th is None:
         return E
 
     # Hysteresis thresholding
+    # Scale the thresholds to the color range of the image
     Tl *= (E.max() - E.min()) / 255
     Th *= (E.max() - E.min()) / 255
     T = zeros(F.shape, dtype=bool)
@@ -64,7 +65,8 @@ def canny(F, s, Tl=None, Th=None):
         E[0, x] = E[F.shape[0] - 1, x] = 0
 
     def follow_nb(y, x):
-        """Follow the neighbouring pixels of an edge pixel in E recursively."""
+        """Follow the neighbouring pixels of an edge pixel in E recursively.
+        Add pixels with a value higher than Tl to the bitmap."""
         if T[y, x]:
             return
 

improc/ass4/report/report.tex

@@ -167,4 +167,46 @@ are separable as well:
 	\end{tabular}
 \end{table}
 
+The separability property is used in the \texttt{gD} function, by calling the
+\texttt{convolve1d} function separately for each direction. This implementation
+yields the 2-jet of the cameraman image in figure \ref{fig:jet}.
+
+\begin{figure}[H]
+	\label{fig:jet}
+	\center
+	\includegraphics[scale=.4]{jet_3.pdf}
+	\caption{The result of \texttt{python gauss.py jet 3}.}
+\end{figure}
+
+\section{Canny Edge detector}
+
+The Canny Edge Detector is implemented in the file \emph{canny.py}. For the
+algorithm, we used the Wiki
+page\footnote{\url{http://en.wikipedia.org/wiki/Canny\_edge\_detector}}. The
+different Wiki sections are marked by comments with similar descriptions. Since
+the Wiki page is self-explanatory, we will not discuss the algorithm itself in
+this report.
+
+The program usage is as follows:
+\begin{verbatim}
+python canny.py SCALE [ LOWER_THRESHOLD HIGHER_THRESHOLD ]
+\end{verbatim}
+The scale is obviously used for finding the intensity gradient, and the
+thresholds are used in the "Hysterisis thresholding" part. Note that the
+thresholds are optional, because the assignment instructs to create a function
+\texttt{canny(F, s)} without any arguments for thresholds. Therefore, we were
+not sure whether to implement this section. If the thresholds are specified,
+the resulting plot will contain an additional image containing a binary image
+of edges. The thresholds can be specified in the range 0-255, they are scaled
+down by the program to match the image's color range.  An example execution of
+edge detection on the cameraman image using a scale of 2, a lower threshold of
+20 and a higher threshold of 60, can be viewed in figure \ref{fig:canny}
+
+\begin{figure}[H]
+	\label{fig:canny}
+	\center
+	\includegraphics[scale=.5]{canny_2_20_60.pdf}
+	\caption{The result of \texttt{python canny.py 2 20 60}.}
+\end{figure}
+
 \end{document}