Click here to start

Table of contents

Can We Include The Third Dimension During Image Mining?

Retrieve An Image

Content-Based Retrieval

Problem Statement

Slide 5

CBIR Methodology

Slide 7

Illustration : Logo Search

Illustration: Arbitrary Query

Limitations

Illustration: Change in View

Slide 12

Illustration: Depth Variation

New Paradigm for CBIR

Application I: Architecture

Cross-ratio

Feature extraction

X-ratio distribution

Similarity measure

Retrieved images using only photometric features

Use of X-ratio

Results for a different query

Retrieval results with geometric and photometric features

Rail track retrieval

Recall and Precision

Retrieval Performance

Application II: Search Arts and Sports Photography

Examples of low DOF images

Typical Retrieval

The Fact

The Method

Difficulty with exponential mixture model

Pixel labeling

Proper Retrieval

Improper Retrieval

Slide 36

Another Example

Results using Blobworld

Precision-recall diagram

Summary

Looking Ahead

Slide 42

Author: Academy office