CITS4241 Visualisation

Visualising 2D Data

In the lectures we have learned that the visualisation techniques employed depend on the category of data we are dealing with and that we need to know the properties of that data set to employ the best visualisation methods.

In this laboratory exercise you will have a large volumetric data set of scalar values, comprising of (x, y, z, f(x,y,z)). You can find this data set in the /cslinx/examples/CITS4241/FullHead directory. The best way to access to files in this directory is to logon to the School's Linux server using your Linux user account. The storage of this data set is as follows,

The data for each value of z are stored in a separate file. Thus, each file, HEADSQ.n, where n varies from 1 to 94, comprises of a set of (x, y, f(x,y)) values.
The data in each file are square, i.e. they are equally sampled in the x and the y directions.
The (x, y) position data are not stored, as they can be determined from the file structure. Clearly, the data set is from a regular sampling; in particular, the distance between samples is fixed and equal in the x- and y-directions.
The scalar attribute value, f(x,y), for each (x,y) position is stored in two (2) bytes with the least significant byte first. This value will therefore be defined over the 16 bits, but the high order bit (bit 15) is always set, and does not contribute to the scalar value. The high order bit is used to indicate the connectivity between slices of data and MUST be ignored for our purposes. Thus, when processing the high resolution data this bit will need to be masked out.

For instance, if the first (least significant) byte was 00001001 and the second (most significant) byte was 10000001 for a given (x,y) position, then the binary representation of f(x,y) is

In your program, f(x,y) should then be computed as

f(x,y) =	byte1 +	(byte1 is the least significant byte, containing 00001001)
	(byte2 & 0x7f) << 8	(byte2 is the most significant byte, containing 10000001)

where 0x7f is the hexadecimal representation of 01111111.
Thus, f(x,y) = 2^0 + 2^3 + 2^8 = 265 (in base 10)

The ordering of the scalar attribute values, f(x,y), in the file is row major.
The size of each file in the data set is 131072 bytes, so there are 256 x 256 different f(x,y) values stored in each file.

The data in each file belong to the E^S₂ category, so there are a number of techniques we can apply. You can guess from each file name that the data are best visualised as an image where the (x, y) values define the pixel positions and the function values f(x, y, z=constant) represent intensity --- colour won't mean much until we define a suitable colour mapping.

The exercise for this laboratory session is to write a program that can read each of these data files and generate an image for each of them. You can then visualise each image by using a simple image analysis package such as xv on the Linux platform.

Further hints:

xv can read a wide variety of image formats so you will have to convert each slice of data to one of these image formats. We will use the PGM image format - in raw bits. The earlier versions of xv accept only 8-bit (unsigned char) raw data for PGM image files; however, the current version of xv accepts both 8-bit and 16-bit (unsigned short) raw data for PGM image files. PGM files begin with some header information (which is interpreted as ASCII) followed by the raw binary pixel data. That is,
```
		P5
		x_size y_size
		maximum_pixel_value
		raw_data...
```
Note that the first line is capital P5. The items in the header are whitespace delimited. Use a newline character to terminate each line.

Example 1: If the data are 64 x 64 and the pixel values range from 0 to 255 then the header might look something like the following:

P5

64 64

255

raw_data...

In this case, xv expects each pixel value to be 8 bits in length, as the maximum value 255 just fits into one single byte (being an unsigned char).

Example 2: If the data are 512 x 512 and the pixel values range from 0 to 599 then the header might look something like the following:

P5

512 512

599

raw_data...

In this case, xv expects each pixel value to be 16 bits in length, as the maximum value 599 requires two bytes to store.

Note: There are no whitespace included in the raw data block. To write the raw data block, use the C function fwrite (or an equivalent Java method). See also the man page for PGM, by typing man pgm under Linux.

For visualisation purpose, it is often desirable to scale the intensity values so that they are within the interval [0,255], with 0 representing black, 255 representing white, and intermediate values representing various shades of gray (i.e. we follow the 8-bit data format as given in example 1). You will therefore need to go through all the pixel positions and compute the maximum intensity value (stored in variable m, say) and then rescale all the intensity values using an assignment statement such as

f(x,y) = f(x,y) * 255.0 / (double)m

Two sample PGM image files, HEADSQ.1.pgm and HEADSQ.2.pgm, have been generated for you. You should inspect the first three lines of these files and view the images via

xv HEADSQ.1.pgm
xv HEADSQ.2.pgm

After generating all the PGM image files, try

xv *.pgm

Click the right button when the mouse is positioned inside the window to bring up a dialog box, then click the "Next" button shown in the dialog box to view the next PGM image and so on.