Fred Park
  • Home
  • Research
  • Teaching
  • Resources
  • Personal
  • Blog
Research Interests: Machine Learning, Computer Vision, Mathematical Image Processing, Numerical Analysis, Convex and Non-Convex Optimization, PDE's, Scientific Computing
Preprints:​
  • ℓ0 Regularized Structured Sparsity Convolutional Neural Networks
  • ​Geodesic Active Contours with Shape Priors for Segmentation, Disocclusion, and Illusory Contour Capture
  • A Weighted Difference of Anisotropic and Isotropic Total Variation for Relaxed Mumford-Shah Image Segmentation
  • Image Segmentation Using Clique Based Shape Prior and the Mumford Shah Functional
  • Parallelization of a Color-Entropy Preprocessed Chan-Vese Model for Face Contour Detection on MultiCore CPU and GPU
  • Robust and Efficient Implicit Surface Reconstruction for Point Clouds Based on Convexified Image Segmentation
  • Feature Identification for Colon Tumor Classification
  • A Fourth Order Dual Method for Staircase Reduction in Texture Extraction and Image Restoration Problems
  • Image Decomposition Combining Staircase Reduction and Texture Extraction
  • Recent Developments in Total Variation Image Restoration
  • Simultaneous Total Variation Image Inpainting and Blind Deconvolution
  • Data Dependent Multiscale Total Variation Based Image Decomposition and Contrast Preserving Denoising
  • Solution Dynamics, Causality, and Critical Behavior of the Regularization Parameter in Total Variation Denoising Problems
Recent Research Snapshots Below:

Tracking example. B2 Stealth bomber is accurately tracked despite regions in background having similar intensities.
Picture
L2 TV Caveats: Isotropic regularization hinders ability to capture chair leg boundary correctly. Increasing regularization yields a loss of features.
Picture
Weighted difference of L1 and L2 TV norms yields a more accurate boundary segmentation. 
Non-Convex Optimization with Applications to Segmentation and Tracking 
Published in:
Proceedings of the 2016 IEEE International Conference on Image Processing (ICIP), Phoenix Arizona, USA, September 25th-28th.
UCLA CAM Report Preprint​
ICIP 2016 Poster

Picture
face image

Picture
automatic face detection

Face Tracking (Facial Contour)
Face Tracking (with ellipse fitting)
Automatic Face Contour Detection and Tracking
(Ongoing Work)

Published in:
Parallel Computing, August 2015,
pp. 28-49.
DOI: 10.1016/j.parco.2015.07.002


Picture








sparse pointcloud

Picture



non-uniform sampled point cloud with holes

Picture
surface reconstruction

Picture
surface reconstruction

Point Cloud Surface Reconstruction via Convexified Image Segmentation


Published in:
Journal of Scientific Computing
February 2013, Volume 54, Issue 2-3, pp 577-602.
DOI: 10.1007/s10915-012-9674-8


Demo Page

Picture
initial contour
Picture
initial contour
Picture
segmentation with shape param = 0.5
Picture
final segmentation
Picture
segmentation without shape
Picture
segmentation with shape param = 1.0 (disocclusion)
Image Segmentation Using Clique Based Shape Prior and the Mumford Shah Functional

Published in:
Proceedings of the 2015 IEEE International Conference on Image Processing (ICIP), Quebec City, Canada.
27-30th of Sept. pp. 4077-4081.
Recognized as part of the “Top 10%” papers in ICIP 2015.
DOI:

10.1109/ICIP.2015.7351572


​Shape Prior Segmentation and Disocclusion Video Demos:

Links:
Whittier College
Whittier College Math 
UCLA Math
UCI Applied Math