Visualization, Spring'98, Yap

[PREVIOUS LECTURE] [NEXT LECTURE] [ALL LECTURES] [CLASS PAGE]

LECTURE 8: ACTIVE VISUALIZATION

  1. INTRODUCTION

    We address the problem of visualization large and complex geometric models. Our approach will try to exploit some basic properties of human vision. This lecture gives some background.

  2. REFERENCES

    We will shortly distribute some papers for your reading.

  3. What is Active Visualization?

    We use this term to describe an interactive computer-aided visualization of a synthetic scene or a computer artifact.
    -- "Interactivity" here implies the user is selecting/suppressing details to be seen, making viewing decisions in both space and time.
    -- We say "active" because the viewer can freely move (body or head) through the model.
    -- But "active" here includes another kind of motion: independent eye motions. Furthermore, we assume that our images will have variable resolution, roughly tracking the eye.
    -- Besides "pure visualization", we include situations where the visualization is combined with other activities or objectives (e.g. navigation, mapping activities, disease diagnosis in medical visualization).
    -- The computer artifact need not be a static model, but could be a time-sequence model.

  4. SCENARIOS OF ACTIVE VISUALIZATION

    • (A) Architectural Walkthroughs:
      -- E.g., a model of an environment or city.
      -- E.g., the interior of a building, submarine or airplane.
    • (B) Browsing and viewing an internet gallery or image archive:
      -- we need access to images of varying scales (thumbnail images to full-blown images, say, 10K by 10K pixels), and across the logical organization of the images in the database.
    • (C) Geographical Information System (GIS):
      -- this include simple map servers.
    • (D) Viewing a movie of a computer simulated phenomena:
      -- E.g., impact of a comet collision, or oil spill.
      -- The user not only wants arbitrary playbacks, but the ability to choose and compose the images.
    • (E) Visualizing volumetric data:
      -- Again, we require "projection" of 3D data to 2D, but these projections are more complicated and varied.
      -- Medical imageries are often based on volumetric data.
    • (F) Artificial Reality
    We are allowed some preprocessing of the models. In some applications, we may allow the user super-human visualization capabilities (e.g., X-ray vision).

  5. ABSTRACT FORMULATION OF PROBLEM

    ``Given a model M (mainly geometric),
        and a class P of projections (i.e., maps from M into images),
        to construct a data structure and algorithms
        which facilitates the rapid formation of any sequence of projections.''
    --Geometric means the model basically lives in Euclidean space (R^2, R^3 typically).
    --The projections P often has a natural topology (e.g., it can be identified with the class of linear maps from R^3 to R^2). "Coherence" of projections can then be exploited.
    --It is useful to separate (as in computer graphics) each projections in P into two parts:
    1. Linear Transformation of Space
    2. Picture composition parameters (such as "camera parameters" and foveation parameters).
    -- Verify that the above scenarios fit this formulation. What is M and P in each case?
    -- The problem can be viewed as a "preprocessing data structure" problem, common in Computational Geometry.

  6. Some Research Issues

    We need to achieve realtime frame rates (say, 15-30 frames per second). We cannot afford to throw the entire model into the graphics rendering pipeline -- even with dedicated hardware and parallel computing power. Some useful techniques include:
    1. Visibility culling
    2. Database management and predictive data motions
    3. Hierarchical level-of-detail (LOD) models
    4. Precomputation of radiosity (or other light models)
    5. Foveated image rendering
    6. Preprocessing for fast view projection
    7. Online scheduling problems (RESERVATION PROBLEMS)

  7. Human Visual System

    Visualization is complex but it is ultimately a psycho-physiological process.
    -- As such, some limits and parameters of the human visual system can be exploited.
    1. The human eye
    2. The Neuro-anatomy of the visual system
    3. Eye Movements

  8. Eye Tracking

     Do we need it? In visualization, the user is motivated to give us the needed information.

  9. Active Vision

  10. Space Variant Images

[PREVIOUS LECTURE] [NEXT LECTURE] [ALL LECTURES] [CLASS PAGE]