mobile image processing

16
Mobile Image Processing Hamed Ordibehesht Mohammad Zand Supervisor: Miroslaw Staron 1

Upload: rasia

Post on 23-Feb-2016

26 views

Category:

Documents


0 download

DESCRIPTION

Mobile Image Processing. Hamed Ordibehesht Mohammad Zand Supervisor: Miroslaw Staron. Overview. Project Description and Assumptions Image Processing Steps Preprocessing BLOB Detection Feature Recognition Efforts Outcomes Further Work. About The Project. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Mobile Image Processing

1

Mobile Image Processing

Hamed OrdibeheshtMohammad Zand

Supervisor: Miroslaw Staron

Page 2: Mobile Image Processing

2

Overview

• Project Description and Assumptions• Image Processing Steps– Preprocessing– BLOB Detection– Feature Recognition

• Efforts• Outcomes• Further Work

Page 3: Mobile Image Processing

3

About The Project• A quick and dirty way of getting early indication of certain

characteristics of the design• Processing Hand-Drawn Class-Diagram– Calculating some simple metrics such as structural complexity in

a dirty way– Impact on quality of the architecture

• Using Symbian Cell-phone• Proof of Concept• Applied IT Project– Solving an existing IT problem by applying scientific findings and

techniques

Page 4: Mobile Image Processing

4

Assumptions

• Consistent drawing style• Rectangular class elements which are big

enough to be recognized as features not noises

• Drawing without textual elements• Using only horizontal and vertical lines

Page 5: Mobile Image Processing

Processing Steps1. Preprocessing• Noise Elimination• Edge detection• Shape refinement

2. BLOB Detection3. Feature Recognition• Domain heuristics

5

Preprocessing

BLOB Detection

Feature Recognition

Page 6: Mobile Image Processing

Preprocessing• Input: digital photo taken by the camera• Noise Elimination by

– Applying symmetric Gaussian lawpass filter• hsize = 15• Sigma = 10• Values through empirical

– Grayscaling– Resizing

• Bicubic Interpolation• Antialiasing• Scale factor = 60%

6

Gaussian Filter Grayscaling Resizing Edge

DetectionShape

Refinement

Page 7: Mobile Image Processing

7

Preprocessing (cont.)

• Edge Detection with– Sobel operator for calculation of threshold value

• Shape Refinement by Morphological operations– Dilation

• Optimal Value = 3• Structuring elements => horizontal and vertical lines

– Closing: combination of Dilation and Erosion• Optimal Value = 5• Structuring Elements => square

• Output: Resampled image

Page 8: Mobile Image Processing

8

Preprocessing Output

Page 9: Mobile Image Processing

BLOB Detection

• Feature Detection– Connected Components– Labeling– Bounding Box calculation

9

Connected Components Labeling Framing

Page 10: Mobile Image Processing

10

BLOB Detection Output

Page 11: Mobile Image Processing

11

Feature Recognition

• Recognition of the diagram elements• Count the number of classes• Process– Assumptions

• Class element minimum bounding box size• Cross lines as

– Domain Heuristics• Class elements do not intersect• A class element’s width ~> height• A Class element consist of maximum two segments which

intersect or align

Page 12: Mobile Image Processing

12

Project Plan

Page 13: Mobile Image Processing

13

Efforts• 580 hours• Reading LOTS of materials

– Research around recent Image Processing Techniques• Learning how to work with MATLAB and Symbian

developing• Developing and comparing some image processing methods

– Blob Detection and Feature Extraction– Noise Elimination– Feature Recognition– Domain Heuristics

Page 14: Mobile Image Processing

14

Outcomes

• Novel noise elimination algorithm• Metrics collection result not accurate enough• Experiencing MATLAB• Symbian development experience• Still at development stage

Page 15: Mobile Image Processing

15

Further Work

• Work on the recognition algorithm for better accuracy

• Development of Symbian application• Run an experiment

Page 16: Mobile Image Processing

16

Thanks, Any Questions

?