14
JCDL 2005 – June 8 th , 2005 User Perceptions of User Perceptions of Digital Image Digital Image Similarity Similarity Unmil Karadkar, Richard Furuta, Jeevan Joseph John Center for the Study of Digital Libraries Texas A&M University Jin-Cheon Na Division of Information Studies Nanyang Technological University

JCDL 2005 – June 8 th, 2005 User Perceptions of Digital Image Similarity Unmil Karadkar, Richard Furuta, Jeevan Joseph John Center for the Study of Digital

  • View
    213

  • Download
    0

Embed Size (px)

Citation preview

JCDL 2005 – June 8th, 2005

User Perceptions of User Perceptions of Digital Image SimilarityDigital Image Similarity

Unmil Karadkar, Richard Furuta, Jeevan Joseph John

Center for the Study of Digital Libraries

Texas A&M University

Jin-Cheon Na

Division of Information Studies

Nanyang Technological University

JCDL 2005 – June 8th, 2005

OutlineOutline

• Motivation

• Study – Research questions– Setting and Design– Analysis of results

• Significance– Theoretical– Practical

JCDL 2005 – June 8th, 2005

MIDAS MIDAS (Multi-device Integrated Dynamic Activity Spaces)(Multi-device Integrated Dynamic Activity Spaces)

Multi-device Integrated Dynamic Activity Spaces

World-WideWeb

P E R C E P T I O N

JCDL 2005 – June 8th, 2005

Context of Information SpaceContext of Information Space

• Images– Graphs (visual data representations)– Line drawings (cartoons, comic strips)– Art (sketches, paintings)– Photographs (scanned or digital)

• Quick growth– Digital cameras– Camera phones

• Text • Audio

• Video • Animation

JCDL 2005 – June 8th, 2005

Research QuestionResearch Question

• How does the variation of image attributes affect user perception of images?

Variables:• Size (scaling)• Colors (change of color depth)

– How does the perception of similarity change?– Do users allow automatic substitution of

modified images for best display on a device?– How does variation affect the confidence of

image substitution?

JCDL 2005 – June 8th, 2005

Image CharacteristicsImage Characteristics

• Size (pixels)– 160x120 (smart phone)– 320x240 (PDA)– 640x480 (CSISD schools)– 800x600 (notebooks)– 1024x768 (desktop computers)

• Colors– 2 (b/w) (1-bit)– 4 (2-bit)– 16 (4-bit)– 256 (8-bit)– 16 Million (24-bit)

• Image types– People – Nature

– Structures – Text

• 4 images of each type

JCDL 2005 – June 8th, 2005

Subject Pool CharacteristicsSubject Pool Characteristics

• 5 Users

• Graduate students and staff

• Male and female

• Different academic backgrounds– Engineering, sciences, architecture

• Different nationalities and cultures

• Mid-twenties to early forties

JCDL 2005 – June 8th, 2005

Setting of the StudySetting of the Study

• Computer with 2 identical displays

• One image on each display

• Time not a factor

• Users could take a break

• Users could clarify doubts and ask questions

• Evaluator did not interrupt the users during the task

JCDL 2005 – June 8th, 2005

Design of StudyDesign of Study

• Users viewed pairs of images

• 3 questions per pair– Perceived level of similarity (9 pt. Likert scale)– Would they allow substitution of one image for

another (9 pt. Likert scale)– Automatic substitution acceptable (Yes/No)

• No confounding effects– Each pair differed in size or colors but not both

• 20 pairs for each variable

JCDL 2005 – June 8th, 2005

Results – Results – Perception of SimilarityPerception of Similarity

• Scales for size and color are not comparable

Steps Within color

Color to grayscale

0 - 6.89

1 6.69 5.57

2 4.6 4.55

3 5.86 4.00

4 4.14 -

Steps Similarity

0 -

1 7.12

2 7.11

3 6.87

4 6.14

Size Color

JCDL 2005 – June 8th, 2005

Results – Similarity by image type Results – Similarity by image type

• People’s faces• Legibility of textual elements• Nature and structural images scale well

Steps nature people structures text

1 6.57 7.17 7.67 6.78

2 7.33 7.09 6.78 7.25

3 7.75 7.00 7.00 6.29

4 7.00 5.71 7.50 5.00

Size

JCDL 2005 – June 8th, 2005

Results – Image Substitution Results – Image Substitution

• Automatic substitution– High confidence but fewer instances

• Substitution with warning– More instances but confidence dropped

Steps Automatic substitution

Confidence Substitution with warning

Confidence

1 44% 7.66 56% 6.05

2 49% 7.42 51% 6.00

3 50% 7.13 50% 6.50

4 21% 7.47 79% 4.18

Size

JCDL 2005 – June 8th, 2005

Significance of ResultsSignificance of Results• Theoretical

– Provides an insight into human image perception

– Nature and structural images scale well– Images of people and text scale to a lesser

degree

• Practical– Design guidelines for image optimization

• Scale images rather than reducing colors

– Use other techniques, such as cropping for better scaling of textual and people images

JCDL 2005 – June 8th, 2005

Questions?