Practical Tutorial 2 (Assessed) - Volume Rendering

The goal of this tutorial is to visualise a head CT dataset by writing your own volume renderer.

This coursework can be done individually or in groups of up to three.

The solutions should be submitted via CATE by midnight Monday 7th November.

This Coursework can be found at:

http://www.doc.ic.ac.uk/~eedwards/teaching/2012/AGV/VolRenCoursework/

Learning outcomes

At the end of this coursework you will

Data

You will find a datasets HeadCT.vtk . This is a vtkStructuredPoints datasets containing a CT scan from the Visible Human Project

There is also a skeleton program VolumeRender.cpp and a corresponding Makefile. If you are using an Ubuntu 10 machine there is an alternative Makefile . To compile the example simply copy these into a directory and type "make". If you want to amend the Makefile to use another version of VTK that's fine. To run the program type "volumeRender HeadCT.vtk tmp.png". This renders a single slice from the dataset on the screen and saves this rendering to tmp.png.

The scalar values in the data range from 0 to around 2500. You will need to explore the data to discover which values correspond to which tissues. You can use the subsampled version for (HeadCTsub.vtk) for experimentation but the final images should come from the full dataset HeadCT.vtk.

The file VolumeRender.cpp contains 3 functions. A "main" which sets up the display to render the 3D data into a 2D RGB image (range 0.0-1.0 for each colour). You should not need to amend the "main".

The TransferFunction takes the short data from the 3D image and converts this to an rgb_alpha value. In this case it simply copies the range of greyscales in the image to 0.0-1.0 in the rgb image. You will need to adjust this TransferFunction to achieve different renderings in the exercises below.

The VolumeRender function loops through the CT data and copies slice 200 into the RGB image.

Exercise 1 - Maximum intensity projection

Amend the VolumeRender function so that it loops through the y voxels. This is effectively parallel projection along the y-axis and ray accumulation takes place for every voxel. Write a new generic "Composite" function that takes a current ray colour and combines it with a value from the current voxel. In this instance, make the Composite function provide the maximum value along the ray. This should provide a rendering showing the skull and highlighting the fillings in the teeth.

Exercise 2 - "X-ray" projection

Alter the Composite function so that it implements the over operator. Amend the transfer function so that the colour is always white, but the alpha value starts at zero for some value and ramps up linearly with increasing scalar value. The maximum value for the given dataset should still be quite small to avoid whiting out the rendering. Experiment with the values to provide a rendering that looks similar to an X-ray.

Exercise 3 - Face and skull rendering

Change the transfer function so that it provides a reasonable skin tone and so that the alpha value ramps up from 0.0 to 1.0 over a small range at the value corresponding to skin. The rendering will appear quite flat but should cover the area of the face. Choose another threshold to provide a similar rendering of the skull. The flat nature of these renderings underlines the need to incorporate lighting.

Exercise 4 - Face and skull rendering with lighting

To provide lighting you will need to implement volumetric gradients as described in the notes. For each voxel, calculate the gradient by finite difference. As we are rendering in the y direction you should scale the voxel colour by the dot product of the normalised gradient with (0,1,0). A few things to note are that you should use the absolute gradient and that you should only provide lighting for gradients where the magnitude exceeds a threshold. The gradient direction is not well approximated for small gradient magnitudes. This method should produce decent renderings of the face and the skin.

Exercise 5 - Combined face and skull rendering with lighting

Incorporating volumetric lighting, alter the transfer function to include a spike at the level of the skin that reaches a smaller alpha than 1.0 and the same ramp for the skull as exercise 4. This should produce a rendering showing the skull through transparent skin.

Notes

Please incorporate all the images produced into a single document. You should submit this along with your source code VolumeRender.cpp and include a description of what thresholds and opacities were used so these images can be reproduced. The source file should include all the code to run any of the exercises (commented out if necessary).

The VTK window includes an interactor that allows you to alter the thresholds visible in the window by dragging the mouse. This is useful for debugging purposes, but the PNG image is saved before any interaction and you should leave it functioning in this way. Your code, not the interaction, should produce the given result. Also, VTK includes many helper functions for volume rendering. You should not use these - the aim of the exercise is to write the volume renderer yourself. This is quite a noisy CT scan so don't expect perfect results.

YOU CAN DO THIS PROJECT INDIVIDUALLY OR IN GROUPS OF UP TO THREE

SUBMIT THE TWO FILES ON CATE - VolumeRender.cpp and report.pdf by MIDNIGHT MONDAY NOVEMBER 7TH

Philip Edwards
Last modified: Fri Oct 21 02:23:59 BST 2011