Implementation & Scheduling

Milestone 0 - Initial Web Site & Planning: Mar 11 - 17, 2006

• Determine what noise models to implement [done]

• Determine what filters to implement [done]

• Create initial web site [done]

• Post project outline [done]

• Ensure vuVolume compiles on Suse 10 (for visualizing datasets later) [done]

Milestone 1 - Stage 1 - Median Filter: Mar 18 - 24, 2006

• Adapt existing code for an analytical dataset: [done]

  • Marschner Lobb

• Research & implement salt & pepper noise model [done]

• Implement default noise error metric [delayed]

  • Root of mean squared differences at sample locations

• Discuss candidate error metrics for measuring effect of noise [done]

  • Perhaps a good metric should be a function of both the mean and variance of the per-sample error.

• Sample the ML with salt & pepper noise [done]

  • CC 100%

  • BCC 71%

    • By this I mean the BCC dataset contains 29% less samples than the CC counterpart.

• Implement spherical median filter for Cartesian (CC) grid [delayed]

  • Fixed filter radius for now

• Implement spherical median filter for BCC grid [delayed]

  • Fixed filter radius for now

• Render the noisy CC / BCC datasets, their de-noised and noise-free counterparts in vuVolume to ensure correctness of the salt & pepper noise model, and the correctness of the median filter. [partly delayed]

• Work on parts of the final write up [done]

Milestone 2 - Stage 1 - Gaussian Smoothing: Mar 25 - 31, 2006

• Research & implement Gaussian white noise model [done]

  • Is Gaussian noise and white noise the same thing?

• Sample the ML with Gaussian noise [done]

  • CC 100%

  • BCC 71%

• Implement spherical median filter for Cartesian (CC) grid [partly delayed]

  • Fixed filter radius for now

• Implement spherical median filter for BCC grid [partly delayed]

  • Fixed filter radius for now

• Implement spherical Gaussian smoothing filter for Cartesian (CC) grid [partly delayed]

  • Fixed filter radius for now

• Implement spherical Gaussian smoothing filter for BCC grid [partly delayed]

  • Fixed filter radius for now

• Render one set of image for CC, and another for BCC [partly delayed]

  • Noisy ML dataset

  • De-noised ML dataset

  • Noise-free ML dataset

• Determine the error metric(s) to use [done]

• Figure out how to grow a neighbourhood in CC, BCC [done]

• Work on parts of the final write up [done]

Milestone 3 - Stage 2 - Comparison of CC / BCC : Apr 1 - 7, 2006

• [In view: deadlines this week]

  • 815 paper presentation

  • Paper reading presentation (potentially)

  • 815 all assignments are due

• Implement spherical median filter for Cartesian (CC) grid  [done]

• Implement spherical median filter for BCC grid  [done]

• Implement spherical Gaussian smoothing filter for Cartesian (CC) grid [partly delayed]

• Implement spherical Gaussian smoothing filter for BCC grid [partly delayed]

• Render one set of image for CC, and another for BCC [partly delayed]

  • Noisy ML dataset

  • De-noised ML dataset

  • Noise-free ML dataset

• Grow a neighbourhood in CC [partly delayed]

  • Record radius of corresponding continuous filters

  • Record the number of samples in neighbourhood

  • Treat (radius, num_samples) as a 2D point

  • Plot the 2D points for various-sized neighbourhoods, and connect them linearly

• Repeat this neighbourhood growth for BCC [delayed]

• Overlay the two neighbourhood growth plots for CC and BCC [delayed]

  • Choose some set of radii (of the continuous filter) to compare de-noising on CC and BCC

• For each radius in the set, for CC ML  [delayed]

  • Measure error due to noise after the de-noising step

  • Treat (radius, num_samples, error) as a 3D point

    • Could also treat (radius, erro) as a 2D point

  • Plot the 3D (or 2D) points for all radii chosen

    • Could use more than one error metric

    • If using multiple error metrics, produce one plot per metric

• Repeat for BCC ML (i.e. produce plot)  [delayed]

• Overlay the corresponding CC/BCC error plots  [delayed]

• Formulate conclusions  [delayed]

• Finalize write up  [delayed]

  • Were expected results observed?

  • Were there any problems encountered?

Milestone 4 - PowerPoint Presentation: Apr 8 - 11, 2006

• [In view: deadlines this week]

  • 815 all assignments are due April 18

• Implement spherical Gaussian smoothing filter for Cartesian (CC) grid [done]

• Implement spherical Gaussian smoothing filter for BCC grid [done]

• Render one set of image for CC, and another for BCC [done]

  • Noisy ML dataset

  • De-noised ML dataset

  • Noise-free ML dataset

• Implement default noise error metric [inserted, done]

  • Variance

• Check that noise content is comparable in CC/BCC pairs. [inserted, done]

  • For salt & pepper, check percentage of corruption.

  • For Gaussian: check variance of noise.

• Grow a neighbourhood in CC [done]

  • Record radius of corresponding continuous filters

  • Record the number of samples in neighbourhood

  • Treat (radius, num_samples) as a 2D point

  • Plot the 2D points for various-sized neighbourhoods, and connect them linearly

• Repeat this neighbourhood growth for BCC [done]

• Overlay the two neighbourhood growth plots for CC and BCC [done]

  • Choose some set of radii (of the continuous filter) to compare de-noising on CC and BCC

• For each radius in the set, for CC ML [done]

  • Measure error due to noise after the de-noising step

  • Treat (radius, num_samples, error) as a 3D point

    • Could also treat (radius, erro) as a 2D point

  • Plot the 3D (or 2D) points for all radii chosen

    • Could use more than one error metric

    • If using multiple error metrics, produce one plot per metric

• Repeat for BCC ML (i.e. produce plot) [done]

• Overlay the corresponding CC/BCC error plots [done]

• Formulate conclusions  [done]

• Finalize write up [done]

  • Were expected results observed?

  • Were there any problems encountered?

• Bonus: produce CC/BCC neighbourhood plot in Matlab [inserted, done]

• Produce Doxygen documentation for code. [inserted, done]

• Ensure references are complete. [inserted, done]

• Update and finalize project web site.  [inserted, done]

• Prepare a 20 page presentation. [done]

• Prepare possible questions and answers.  [done]

• Prepare outline for a demo. [done]

Demo & Presentation: Apr 13 3:40 - 4:20pm, 2006

• In the MTF

• [MTF closes next week, so no extensions given]