Indexing Still and Moving Images



                   Ian Davis
                   Global Taxonomy Delivery Manager
     ...
Indexing Still and Moving Images


  Still image indexing
  Moving image indexing

  Two content types can often be treate...
Who uses still images?

           Newspaper and magazine publishers
           Book publishers
           Internet or ...
What do People Want?


         Access to images based on attributes:

         e.g. images photographed at a specific tim...
What do People Want?


         Access to images based on specific depictions:

         e.g. 'sun loungers on a tropical ...
What do People Want?


         Access to images based on conceptual terms:

         e.g. 'eerie images‘

         e.g. '...
What do People Want?


           Access to images based on basic inherent content:

            Colours

            Te...
Fitting the system to the users



   Image indexing should focus on the images and
    also on the people who will be lo...
Fitting the system to the users

       Editorial users often appreciate:
           Full captions of two to three senten...
Indexing Strategy



  Choose from:

   Manual approach
   Automatic approach
   Combined approach




© Copyright 2008...
Fitting the system to the users

   Understand client needs

   Construct an approach to meet these needs

   Choose fr...
Methods of Image Classification and Retrieval


  Semantic content – textual classification and
    retrieval
      Human...
Manual Image Indexing


         Meta-date Coverage – layers of information

               Technical image information
 ...
Manual Image Indexing


           Technical image information:

                 Unique scanning numbers
              ...
Manual Image Indexing


              Basic Image Attributes:

                    Photographer name
                   ...
Manual Image Indexing


    Image Content Information:

     Actions and Events
     Animals and Plants
     People: ro...
Manual Image Indexing


    Image context/aboutness information:

     Emotions
     Abstract concepts




© Copyright 2...
Automatic Image Indexing


    Inherent in image pixels:

     Colours
     Textures
     Simple shapes




© Copyright...
Video Indexing

  Layers of information
   Indexing the whole piece
   Indexing the scene
   Indexing the frame

  Type...
Video Indexing

   Dealing with video includes all the challenges of
    still image indexing

   Plus the technical iss...
Video Indexing

   Much initial textual and audio data can be
    captured and indexed automatically
   System can be tr...
Thank You!

                      ian.davis@dowjones.com



© Copyright 2008 Dow Jones and Company                | 22
Upcoming SlideShare
Loading in...5
×

Indexing Still and Moving Images

840

Published on

This presentation outlines the ways in which still images and video can be categorised to boost findability.

Published in: Business, Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
840
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
0
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Transcript of "Indexing Still and Moving Images"

  1. 1. Indexing Still and Moving Images Ian Davis Global Taxonomy Delivery Manager Dow Jones Consulting Services © Copyright 2008 Dow Jones and Company | 1
  2. 2. Indexing Still and Moving Images Still image indexing Moving image indexing Two content types can often be treated very similarly Two content types also bring unique challenges © Copyright 2008 Dow Jones and Company | 2
  3. 3. Who uses still images?  Newspaper and magazine publishers  Book publishers  Internet or Intranet sites  Music labels  Advertising agencies  Design houses  Academics  Corporates  TV companies © Copyright 2008 Dow Jones and Company | 3
  4. 4. What do People Want? Access to images based on attributes: e.g. images photographed at a specific time. e.g. images photographed at a specific place. e.g. images created using a certain photographic process. © Copyright 2008 Dow Jones and Company | 4
  5. 5. What do People Want? Access to images based on specific depictions: e.g. 'sun loungers on a tropical beach‘ e.g. 'peacock, full tail display and head’ e.g. 'individuals or couples in modern, clean, bright kitchens or living rooms - ideally to show blur of motion.' © Copyright 2008 Dow Jones and Company | 5
  6. 6. What do People Want? Access to images based on conceptual terms: e.g. 'eerie images‘ e.g. 'exciting, speedy images’ e.g. 'images illustrating individuality' © Copyright 2008 Dow Jones and Company | 6
  7. 7. What do People Want? Access to images based on basic inherent content:  Colours  Textures  Shapes © Copyright 2008 Dow Jones and Company | 7
  8. 8. Fitting the system to the users  Image indexing should focus on the images and also on the people who will be looking for them. How images are be indexed should relate closely to this. © Copyright 2008 Dow Jones and Company | 8
  9. 9. Fitting the system to the users Editorial users often appreciate:  Full captions of two to three sentences  Detailed factual information Ad and design users often:  Care little for concrete facts  Like to browse for ideas  Like to search using concepts © Copyright 2008 Dow Jones and Company | 9
  10. 10. Indexing Strategy Choose from:  Manual approach  Automatic approach  Combined approach © Copyright 2008 Dow Jones and Company | 10
  11. 11. Fitting the system to the users  Understand client needs  Construct an approach to meet these needs  Choose from the available techniques and technologies  Consider textual or CBIR, or even textual and CBIR © Copyright 2008 Dow Jones and Company | 11
  12. 12. Methods of Image Classification and Retrieval Semantic content – textual classification and retrieval  Human analysis of images  Metadata development  Thesauri or ontology supported classification and retrieval Content based image retrieval (CBIR)  Automatic analysis of low level image attributes © Copyright 2008 Dow Jones and Company | 12
  13. 13. Manual Image Indexing Meta-date Coverage – layers of information  Technical image information  Basic image attributes  Image content information  Image context/aboutness information © Copyright 2008 Dow Jones and Company | 13
  14. 14. Manual Image Indexing Technical image information:  Unique scanning numbers  Scanner type  Processor name  Processing activities  Batch numbers  Sizes of image files © Copyright 2008 Dow Jones and Company | 14
  15. 15. Manual Image Indexing Basic Image Attributes:  Photographer name  Photography Types  Location photographed  Date photographed  Work type  Points of view © Copyright 2008 Dow Jones and Company | 15
  16. 16. Manual Image Indexing Image Content Information:  Actions and Events  Animals and Plants  People: roles, occupations, ethnicity, attributes, and anatomy  Topography  Built works  Things © Copyright 2008 Dow Jones and Company | 16
  17. 17. Manual Image Indexing Image context/aboutness information:  Emotions  Abstract concepts © Copyright 2008 Dow Jones and Company | 17
  18. 18. Automatic Image Indexing Inherent in image pixels:  Colours  Textures  Simple shapes © Copyright 2008 Dow Jones and Company | 18
  19. 19. Video Indexing Layers of information  Indexing the whole piece  Indexing the scene  Indexing the frame Types of information  Image information  Textual information  Audio information © Copyright 2008 Dow Jones and Company | 19
  20. 20. Video Indexing  Dealing with video includes all the challenges of still image indexing  Plus the technical issues of capturing textual data about the content  And the issue of automatically identifying scenes and frames  And the issue of capturing, indexing and linking audio soundtracks to these scenes and frames. © Copyright 2008 Dow Jones and Company | 20
  21. 21. Video Indexing  Much initial textual and audio data can be captured and indexed automatically  System can be trained to identify scene breaks and key frames can be selected  Focus is often on improving and monitoring the automated aspects and prioritising what work needs to be done manually – usually focusing on key frames/scenes in a piece © Copyright 2008 Dow Jones and Company | 21
  22. 22. Thank You! ian.davis@dowjones.com © Copyright 2008 Dow Jones and Company | 22

×