Data Structures

It is obvious that the main structures are object-instances of various classes. This classes have implemented some methods, where the program functionality is stored. Here is a short overview of our classes:

• MaskSmoothing (public) - The ''main'' class. Actually it calls the AppFrame class and starts the application.
• AppFrame - This class is a subclass of the already implemented java.awt.Frame class. Here the GUI is implemented, reading and writing the data set. Therefore, all the data sets are finally stored in this class. Also the dominant part of the data visualization is implemented here.
• CanvasSlice, CanvasPerspective - Classes derived from the java.awt.Canvas. These are also used for the data visualization. CanvasSlice simply flushes the input image to the canvas. CanvasPerspective shows a box and three rectangles that represent the top, front and side slice position within the volume.
• AppFrame_AboutBox, AppFrame_SettingsBox, AppFrame_StatisticsBox - Classes of the about, settings and statistics dialog.
• ComplexNumber - One of the things, that are in Java not yet implemented, is the complex computing. Therefore we had to implement it ourselves, to perform mathematical operations in the frequency domain. There are not all of mathematical operations implemented in this class, we have implemented just those methods we needed for the calculation.
• FilterKernel - The class creates a filter according to the settings dialog.
• FilterMaskSmoothing - This is the core of the algorithm calculation. There are FFT, convolution, deconvolution and gradient estimation implemented in this class. For data representation we use one, and three dimensional arrays. For specific operations the 1D representation is preferred, where $f[x][y][z]$ is equal to $g[(z*maxY+y)*maxX+x]$. Another problem is, that Java assumes Little-Endian byte order, but our data set was stored in Big-Endian byte order. Unfortunately it is not possible to specify unsigned property of the basic types. Therefore, a short variable is read from the file as follows:
  x=0x7fff&(input.readUnsignedByte()|input.readUnsignedByte()<<8);
  where the input is an object of class DataInputStream, which is used for reading. As standard input type we took Big-Endian short (16 bit integral type). The reason is, the output of a CT contains also short integers although only 12 bits are used.