Almon David Ing

I am a graduate researcher affiliated with the Center for Perceptual Systems and Department of Psychology at the University of Texas at Austin.  I expect to receive a PhD in August of 2009.

My interests are in vision, perception, neuroscience, and classification modeling.  I was trained by Bill Geisler to study natural image statistics and to run psychophysical experiments.  I studied statistics with Larry Cormack, cognition and speech perception with Randy Diehl, categorization with Todd Maddox, and the neurobiology of working memory with Greg Ashby.

My PhD dissertation focuses on Natural Scene Statistics and the Leaf Segmentation Problem.  The work is an analysis of foliage-rich images. In particular, I studied how properties of these images can be exploited to serve image segmentation.  This work has implications for understanding the nature of human and ideal (computer) visual systems.

As an engineering challenge, the Leaf Segmentation Problem is extremely difficult and shares many properties with other image segmentation problems. The problem's solution is interesting because it may help solve many other image segmentation problems. 

Careful study of this problem also helps answer many general questions about the human visual system. Organic perceptual systems tend to evolve in response to physical properties of natural environments.  Like many primate visual systems, the human visual system is strongly influenced by the properties of these environments.  Therefore any attempt to understand the functionality of these systems should incorporate the study of these environments (e.g. foliage-rich scenes).

ObjectParser Software & Databases

The ObjectParser software application allows a user to segment images quickly and accurately because the user can easily pan and zoom the image during segmentation.  The segmentation data can be easily analyzed because it is saved in a text format (XML-based).  Since many researchers conduct analyses using Matlab, I provide Matlab code for reading the data files.

I designed ObjectParser as a way to collect the ground-truth associated with any kind of image segmentation problem.  In a nutshell, the software allows a user to easily parse objects from images by defining polygons to demarcate object surfaces.  The user can fully specify the nature of any occlusion (including the ability to specify foreground-background information).  The user can also trace and classify surface contours (e.g. surface markings, sharp shadow boundaries).

Multivariate Categorization Library

Many problems in scene statistics can be decomposed into an arrangement of sub-problems.  Many of these sub-problems are classification problems.  An ideal classification system should be able to estimate the probability of category membership given an unclassified stimulus coordinate.  Determining ideal classification performance is extremely complicated.

I developed the Multivariate Categorization Library as an aid for determining ideal classification performance.  Currently I am distributing this library as a collection of Matlab functions.  The library makes it easy to visualize multivariate data, find optimal decision functions (using modified quadratic and kernel density methods), and plot these decision functions with the data.  Simple examples are provided to make it easy for a user to learn how the library works.

Skills & Experience

As an undergraduate at UCSB, I began as an engineering student in the computer science program. I successfully completed many classes in introductory computer science and nearly completed a minor in Business Accounting. Motivated by a desire to understand the computational architecture of the human brain, I eventually chose to pursue a degree in psychology.

In graduate school I completed coursework in statistics, neuroscience, vision, perception, speech perception, and cognition.  I have been a paid research assistant since beginning graduate school, so I feel very comfortable in a productive research setting. I would also feel comfortable teaching the following courses at a top-tier university: Perception (vision, audition, and speech), traditional inferential statistics, modern statistics (using algorithms), computational neuroscience (a.k.a. functional neuroanatomy), object-oriented programming using Java or C#, application development using C# (especially standalone user-interface applications), classification theory, signal detection theory, and general data analysis using Matlab.

My laboratory skills include: Classifier design and implementation, color calibration of cameras and displays, design of visual psychophysics experiments, software design (standalone user-interface applications), object-oriented programming, empirical design and analysis, numerical algorithm design, traditional statistics, modern statistics (using algorithms), natural scene statistics, linear systems, signal detection theory, signal processing, and personnel management.

Presentations & Publications

Ing, A.D.  (in preparation)  PhD Dissertation.

Ing, A.D.  (in preparation)  Optimal classification and real analysis.

Ing, A.D. & Geisler, W.S. (in preparation)  Contour classification statistics for leaf segmentation.

Ing, A.D. & Geisler, W.S. (in preparation)  Foreground-background statistics for leaf segmentation.

Ing, A.D. & Geisler, W.S. (in preparation)  Patch pair statistics for leaf segmentation.

Ing, A.D. & Geisler, W.S.  (2008)  Natural categorization statistics.  Lecture to Cognition & Perception area given October 3, 2008.   [pptx]

Ing, A.D., & Geisler, W.S. (2008)  Natural image statistics and the problem of leaf segmentation.  NETI 2008 Workshop Poster.  [pdf]

Ing, A.D., & Geisler, W.S. (2008) Patch pair statistics for leaf segmentation [Abstract]. Journal of Vision, 8(6):69, 69a, http://journalofvision.org/8/6/69/, doi:10.1167/8.6.69. [pptx] Lecture presented at the annual meeting of the Vision Science Society.

Ing, A.D., & Geisler, W.S. (2006) Ribbon analysis of contours in natural images [Abstract]. Journal of Vision, 6(6):103, 103a, http://journalofvision.org/6/6/103/, doi:10.1167/6.6.103. [ppt] Lecture presented at the annual meeting of the Vision Science Society.

More...

Contributions to Wikipedia

I started the article on Scene Statistics using excerpts from my PhD Dissertation proposal.

I started the article on Ideal Observer Analysis.