Nara Womens University Automatic Viewpoint Selection for a

Nara Women’s University Automatic Viewpoint Selection for a Visualization I/F in a PSE Machiko Nakagawa, Masami Takata, Kazuki Joe Nara Women’s University

Outline Background Explain the Viewpoint Entropy Proposal of View Potential Experiment Discussions Conclusions & Future work Nara Women’s University

Background Importance to select good viewpoints Problem of viewpoint selection Complex object Large-scale data a lot of visualized information huge calculation cost of rendering no criteria for good view ｙ-axis x-axis ？ z-axis data of enough knowledge difficult to Need select good viewpoints time data & visualization technique etc. Nara Women’s University ？

View Selection in PSE Possible visualization without expertise in PSE. View selection by user Eager of automatic viewpoint selection possibility of easier visualization Technique of Automatic Viewpoint Selection with versatility Nara Women’s University

Definition of Good Views No common definition Local definitions depending on each purpose Necessary information　→　visibility Unnecessary information　→　invisibility Good View information NEED USELESS Nara Women’s University

Previous Works Vázquez, “Viewpoint Selection Using Viewpoint Entropy“(2001) A viewpoint definition by information theory Shannon’s Entropy Viewpoint Entropy projected Area the number of visible faces Nara Women’s University

Viewpoint Entropy Nf：the number of faces of the scene Ai：projected area of a face i A 0：projected area of the background in open scenes At：the total projected area of the scene Nara Women’s University

Re-experiment of Viewpoint Entropy (1/2) RE-1 projected area is moved. The number of visible faces is constant best view Movement of a camera As the projected area increases, Viewpoint Entropy increases Nara Women’s University

RE-2 Re-experiment of Viewpoint Entropy(2/2) The number of visible faces is increased. The projected area is almost same as the previous experiment best view As the number of visible faces increases, Viewpoint Entropy increases Nara Women’s University

A Problem of Viewpoint Entropy The same Viewpoint Entropy value Difference in information of views Nara Women’s University

Improvement of Viewpoint Entropy Only two properties for viewpoint selection No other properties which should be Brightness, Color, etc. problems of Viewpoint Entropy plural properties to obtain better views Improvement of evaluation method View Potential Nara Women’s University

Proposal of View Potential W 0： projected area　& the number of visible faces W 1：luminance W 2：chrominance W 3：weight of objects Nara Women’s University

W 1: Luminance(1/2) Brightness is more sensitive than color difference for human perception Dark picture EX) Dark place and/or very small object Bright picture Recognize shape(brightness) Unrecognize color difference Luminance is important for scene Nara Women’s University recognition.

W 1: Luminance(2/2) Calculation of viewpoint selection with view luminance YIQ Color System　 【Y(Luminance ）, I & Q(Chrominance ）】 convert RGB into YIQ Y = 0. 2990 * R + 0. 5870 * G + 0. 1140 * B I = 0. 5959 * R - 0. 2750 * G + 0. 3210 * B Q = 0. 2065 * R - 0. 4969 * G - 0. 2904 * B Nara Women’s University

Luminance Property What’s a good view in luminance ? The value of luminance diffuses. Larger dispersion in luminance should be selected. Nara Women’s University

W 2: chrominance cognition is difference in hue red-green yellow-blue chrominance in data Easy Difficult RGB Color System chrominance in perception different impressions by color mapping bury the difference of color recognition! L*a*b*Color System Nara Women’s University

Chrominance Property Views with higher space frequency are more recognizable. The use of a differentiation filter edge Nara Women’s University

W 3: Weight objects Weight each object as the importance degree The weight of unnecessary objects is 0 Reduction of calculation cost weight: 2 Need No Need weight: 1 weight: 0 Nara Women’s University

Visualization Pipeline (1/2) BYU Data Create Scene vtk. BYUReader vtk. Cube. Source vtk. Poly. Data. Nomals * Generate a Scene * vtk. Poly. Data. Mapper The polygon object is set up vtk. Actor in vtk. Render. Window vtk. Poly. Data. Mapper 3 DS Data vtk 3 DSImporter vtk. Render. Window Nara Women’s University

Pipeline of visualization(2/2) vtk. Render. Window Actor. List vtk. Actor. Collection vtk. Actor vtk. Triangle. Filter vtk. Mass. Entropy Implemented library Get. Information Calculate Entropy take out an Actor of the scene. calculate NULL each To use object. vtk. Mass. Entropy the cell of the polygon is calculate normalized. information Nara Women’s University

vtk. Mass. Entropy Functions vtk. Mass. Entropy Set. Input(vtk. Poly. Data); Input data necessary for calculation Set. Actors(vtk Actor. Collection) Set. Weight(int) Input the weight of each object Get. Entropy() Calculate the Viewpoint Entropy Get. Chromi() Calculate chrominance Get. Cont(vkt. Renderer) Calculate contrast Nara Women’s University

RE-3 Viewpoint Entropy+ Luminance Add the property of brightness to RE-1 Select asymmetry and a contrasty view entropy entrpy＋luminance Nara Women’s University

RE-4 Experiment of Chrominance ： Data description ECMWF The European Center for Medium-range Weather Forecasts provide temperature data of the atmosphere. （1991/1/1) ・Height：Latitude ・Width：Longitude ・time：altitude ・color：temperature Nara Women’s University

Comparison of images from experiment results (1/2) High appraisal Large deviation Low appraisal Small deviation Nara Women’s University

Comparison of images from experiment result (2/2) High appraisal Low appraisal Almost same by human vision Nara Women’s University

Change hue Complex temperature change High appraisal Simple temperature change Low appraisal The impression changes by hue Nara Women’s University

Weighting Objects: Environment RE-5 A scene with several objects A camera moves with a constant distance around the focus point. Nara Women’s University

Weighting Objects no weighting set a value to this object Nara Women’s University

View. Set Change the coefficients of each property A set of good viewpoints Nara Women’s University

Discussions luminance calculating the contrast of the whole scene, The detail of an object might not be presented. improvement by the information of color difference chrominance Not only the chrominance values but also the chrominance degree based on human perception application of texture mapping etc. Nara Women’s University

Conclusions An automatic and general viewpoint selection technique is proposed. View Potential with plural properties is defined. Experiments with some scenes, and selection of good views Nara Women’s University

Future works reduction of calculate cost CPU GPU use of general purpose shade pipes calculate vtk library →　direct. X or Open. GL decrease the number of calculating points. How to move camera Appropriate coefficient for each property by GUI Nara Women’s University