CMSC 828S/838S : Information Visualization

Application Project :
Personal Digital Photo Library

Hyunmo Kang
October 7, 1999

Domain Database : HCIL Digital Photo Database

Visualization Tool : Spotfire pro 4.0

1. Introduction

Recently, as a result of growing popularity of the Internet and digital input devices such as scanners and digital cameras, general computer users happen to have lots of chances to get in touch with all kinds of digital images. So, there are significant challenges to be overcome to effectively organize and access this image information. But according to several user researchs, the organization and retrieval of digital images is a source of great frustration to general users.
In this application presentation, I am going to suggest and derive a new approach to making digital image libraries more easily and efficiently accessible to general users by 1) analyzing the structures of databases and statistics of their contents, 2) finding the interesting features and relations which exist particularly in this kind of databases, and finally 3) analyzing the advantages and restrictions of current existing visualization tools

2. Analysis of Photo-library Database

-  Strucure of Photo-library Database :

• DataBase file : photodb.mdb (MS-access file) - relational database

• PhotoDB Table (Primary Table) :

• Photo ID : Integer (Serial Number) -> primary key (common field)
• Date : Date format (DD-MM-YY)
• Event : Text string
• Location : Text string
• Person : Text string
• Keywords : Text string
• Rating : Integer (1-10) 1:lowest and 10:highest
• Picture : OLE object (jpg file)
• Color : Text string (Color / B/W )
• Locale : Text string (Indoor/Outdoor)
• Contents : Text string (Landscape, Portrait still, Portrait snapshot, Group still, Group snapshot)
• Hyperlink : Hyperlink (The name of the files and directories - relative path)
• Number of People : Integer (greater than or equal to 0)

• Person Table (Related Table) :

• ID : Integer (Serial Number) -> primary key
• Photo ID : Integer  -> foreign key (common field)
• Person : Text string (Name of people in the picture)

-  Statistics & Distribution of Data :

• Distribution of date

• Distribution of location

• Distribution of the number of people per picture

• Distribution of picture rating

• Rate of Color / B/W

• Rate of Contents (Portrait, Group, Landscape)

• Rate of Locale (Indoor, Outdoor)

• Missing data

As databases in general do not always have all data fully defined, photo-library database also has lots of missing data

• Date : 21 out of 100
• Event : 9 out of 100
• Location : 16 out of 100
• Person : 5 out of 100 (landscape : 2)
• Keywords : 31 out of 100
• Rating : none
• Color : none
• Locale : none
• Contents : none
• Number of People : none
  

• Discord between person and photoID attributes

Because the primary key "photoID" from the PhotoDB table is also used as a foreign key in the Person table, the "photoID" field as a foreign key in the Person table does not have unique values. This  may cause a discord problem in visualizing the Photo-library Database on 2-D space in accordance with the fact that the focus is on the picture or on the person.

• Total sum of the number of people per picture (100 pictures) : 351 (average : 3.5 persons / picture)
• Total sum of the number of identified people per picture (100 pictures) : 197 (only 56.1 % of people identified)
• Total number of identified people in the 100 pictures : 69 (no duplicate)

• The number of pictures per (identified) person

• The number of persons per picture

3. Data Visualization via Spotfire

-  The Matrix of Attributes

    Because the spotfire is a 2-D visualization tool, the attibutes matrix is useful to test and find the relationship between every two attributes. The experiments have been performed on only the half of the attributes matrix, because the order of X and Y axis is not so important to find the relation between two attributes. Besides, color and size coding scheme has been used accordingly to represent more than two-dimensional space or to show the relationship between two attributes more clearly.

• x   - untested element
• O  - Interesting feature / strong relation
• ?   - no relation / ambiguous relation  

4. Critique and Suggested Data Visualization

-  Critique for Spotfire

• Good for analyzing data, but not good for searching and browsing data
• The effect of combining slider filters is strictly conjunctive
• 2-D scatterplot representation of digital photo-library database is not so much meaningful in searching the photos
• Hard to see the overview of data distribution for each attribute simultaneously
• Hard to deal with the missing data for each attribute
• No way to browse the images for searching the intended picture

-  Main Issues of suggested visualization technique

• How to handle missing data for each attribute
• How to make use of the full power of Boolean queries
• How to make use of the advantages of Dynamic Queries (Spotfire)
• How to make use of the advantages of Query Preview
• How to make it easy to browse as well as search the pictures
• How to make connection between annotation techniques and search & browse techniques

• Suggested Design for Personal Digital Photo Library Application

• Suggested Design for Query Dialog

• Suggested Design for Photo Detail Viewer

5. References

1. Allan Kuchinsky, Celine Pering, Michael L. Creech, Dennis Freeze, Bill Serra, Jacek Gwizdka.
     "FotoFile: A Consumer Multimedia Organization and Retrieval System", Proceedings of CHI99, pp .496-503, May, 1999.
2. Stephan Greene, Gary Marchionini, Catherine Plaisant, and Ben Shneiderman.
    "Previews and Overviews in Digital Libraries: Designing Surrogates to Support Visual Information Seeking", 1997
3. Ken Fishkin, Maureen C. Stone.
    "Enhanced Dynamic queries via Movable Filters", Proceedings of CHI95, pp. 415-420, May, 1995.
4. Lisa Tweedie, Robert Spence, Huw Dawkes and Hua Su.
    "Externalising Abstract Mathematical Models", Readings in Information Visualization, pp. 253-259, 1999.

Written by Hyunmo Kang, Oct 5, 1999

Web Accessibility