Image Indexing and Retrieval 
Rachmat Wahid Saleh Insani, S.Kom 
Multimedia Database Management System - Chapter 6
Objectives 
• Image indexing and retrieval approaches. 
• Image retrieval based on text description. 
• Image Indexing and Retrieval based on features representation 
(color, shape, and texture). 
• Image similarity calculation. 
• Image indexing and retrieval techniques based on compressed 
image data. 
• Other image indexing and retrieval technique. 
• Integrated image retrieval technique. 
Multimedia Database Management System - Chapter 6
Image Indexing And 
Retrieval Approaches 
Multimedia Database Management System - Chapter 6
Image Indexing and Retrieval 
Approaches 
First approach: a set of attributes. 
Multimedia Database Management System - Chapter 6
Image Indexing and Retrieval 
Approaches 
Second approach: An integrated feature-extraction/ 
object-recognition subsystem. 
Multimedia Database Management System - Chapter 6
Image Indexing and Retrieval 
Approaches 
Third approach: image annotation. 
Multimedia Database Management System - Chapter 6
Image Indexing and Retrieval 
Approaches 
Fourth approach: low level image features. 
Multimedia Database Management System - Chapter 6
Image Retrieval Based 
On Text Description 
Multimedia Database Management System - Chapter 6
Text Based Image Retrieval 
Multimedia Database Management System - Chapter 6
Color Based Indexing and 
Retrieval Technique 
• The idea is to retrieve from database image that has perceptually 
similar colour to the user's query image or description. 
• During retrieval, the distance between the histogram of the query 
image and image in the database are measured. 
• A color histogram H(M) is a vector (h1, …, hj, … hn). 
• hj, number of pixel of image M falling into bin j. 
• hn, number of pixel of image M falling into all bin. 
• Bin, is a discrete color combinations. 
Multimedia Database Management System - Chapter 6
Color Based Indexing and 
Retrieval Technique 
The simplest distance between images I and H is the L-1 
metric, defined as 
nΣ 
d(I,H) = | il − hl | 
l=1 
Example: We have three images of 8x8 pixels and each 
pixel is in one of eight colors C1 to C8. Image 1 has 8 
pixels in each of the eight colors, Image 2 has 7 pixels in 
each of colors C1 to C4, and 9 pixels in each of colors C5 
to C8. Image 3 has 2 pixels in each of colors C1 and C2, 
and 10 pixels in each of colors C3 to C8. Which two images 
are most similar and which two images are most different? 
Multimedia Database Management System - Chapter 6
Improvements to the Basic 
Technique of Color Based IR 
• Making use of similarity among colors. 
• Making use of spatial relationships among pixels. 
• Making use of the statistics of color distribution. 
• Better color representation. 
Multimedia Database Management System - Chapter 6
Image Retrieval Based on 
Shape 
• Images are segmented into individual objects. 
• The basic issue is shape representation and similarity measurement 
between shape representations. 
• A good shape representation and similarity measurement for recognition 
and retrieval purposes have important properties: 
- Each shape should have a unique representation, invariant to 
translation, rotation, and scale; 
- Similar shapes should have similar representations so that retrieval can 
be based on distances among shape representations. 
• The similarity measure between shape representations should conform to 
human perception. 
Multimedia Database Management System - Chapter 6
Image Retrieval Based on 
Texture 
• Texture is described by six features: 
- coarseness, opposite to fine; 
- contrast, dynamic range of gray level, ratio of bow areas, 
sharpness of edges, and period of repeating pattern; 
- directionality, element shape and placement; 
- line likeness, shape of a texture element; 
- regularity, variation of an element placement rule; 
- roughness.the texture is rough or smooth. 
Multimedia Database Management System - Chapter 6
Image Indexing and Retrieval 
Based on Compressed Image Data 
• There are three common compression technique 
for image indexing and retrieval: 
- DCT Coefficient 
- Wavelet Coefficient 
- VQ Compressed Data 
Multimedia Database Management System - Chapter 6
Other Techniques 
• Image Retrieval Based on Model-Based 
Compression 
• Image Retrieval Based on Spatial Relationship 
Multimedia Database Management System - Chapter 6
Image Retrieval based on 
Model-based Compression 
• An object, is represented by a mathematical model 
(parameter or mathematical equation). 
• A very little data required for representing these 
parameters and equations, so a very high 
compression can be achieved. 
• Image distance calculated by parameters 
differences. 
Multimedia Database Management System - Chapter 6
Image Retrieval Based on 
Spatial Relationship 
• A spatial relationship, 
specifies how some object is 
located in space in relation 
to some reference object. 
• Example queries, “find 
images containing a sun 
above to a mountain”. 
• Example application, 
Geographical Information 
System (GIS). 
Multimedia Database Management System - Chapter 6
Integrated Image Indexing 
and Retrieval Techniques 
Structured 
attributes Pictorial queries 
are not supported 
Integrated IR 
Techniques + 
relevance 
feedback 
Text-annotation 
Color based High-level 
abstractions in 
images are not 
supported 
Shape based 
Texture based 
Multimedia Database Management System - Chapter 6

Image Indexing and Retrieval

  • 1.
    Image Indexing andRetrieval Rachmat Wahid Saleh Insani, S.Kom Multimedia Database Management System - Chapter 6
  • 2.
    Objectives • Imageindexing and retrieval approaches. • Image retrieval based on text description. • Image Indexing and Retrieval based on features representation (color, shape, and texture). • Image similarity calculation. • Image indexing and retrieval techniques based on compressed image data. • Other image indexing and retrieval technique. • Integrated image retrieval technique. Multimedia Database Management System - Chapter 6
  • 3.
    Image Indexing And Retrieval Approaches Multimedia Database Management System - Chapter 6
  • 4.
    Image Indexing andRetrieval Approaches First approach: a set of attributes. Multimedia Database Management System - Chapter 6
  • 5.
    Image Indexing andRetrieval Approaches Second approach: An integrated feature-extraction/ object-recognition subsystem. Multimedia Database Management System - Chapter 6
  • 6.
    Image Indexing andRetrieval Approaches Third approach: image annotation. Multimedia Database Management System - Chapter 6
  • 7.
    Image Indexing andRetrieval Approaches Fourth approach: low level image features. Multimedia Database Management System - Chapter 6
  • 8.
    Image Retrieval Based On Text Description Multimedia Database Management System - Chapter 6
  • 9.
    Text Based ImageRetrieval Multimedia Database Management System - Chapter 6
  • 10.
    Color Based Indexingand Retrieval Technique • The idea is to retrieve from database image that has perceptually similar colour to the user's query image or description. • During retrieval, the distance between the histogram of the query image and image in the database are measured. • A color histogram H(M) is a vector (h1, …, hj, … hn). • hj, number of pixel of image M falling into bin j. • hn, number of pixel of image M falling into all bin. • Bin, is a discrete color combinations. Multimedia Database Management System - Chapter 6
  • 11.
    Color Based Indexingand Retrieval Technique The simplest distance between images I and H is the L-1 metric, defined as nΣ d(I,H) = | il − hl | l=1 Example: We have three images of 8x8 pixels and each pixel is in one of eight colors C1 to C8. Image 1 has 8 pixels in each of the eight colors, Image 2 has 7 pixels in each of colors C1 to C4, and 9 pixels in each of colors C5 to C8. Image 3 has 2 pixels in each of colors C1 and C2, and 10 pixels in each of colors C3 to C8. Which two images are most similar and which two images are most different? Multimedia Database Management System - Chapter 6
  • 12.
    Improvements to theBasic Technique of Color Based IR • Making use of similarity among colors. • Making use of spatial relationships among pixels. • Making use of the statistics of color distribution. • Better color representation. Multimedia Database Management System - Chapter 6
  • 13.
    Image Retrieval Basedon Shape • Images are segmented into individual objects. • The basic issue is shape representation and similarity measurement between shape representations. • A good shape representation and similarity measurement for recognition and retrieval purposes have important properties: - Each shape should have a unique representation, invariant to translation, rotation, and scale; - Similar shapes should have similar representations so that retrieval can be based on distances among shape representations. • The similarity measure between shape representations should conform to human perception. Multimedia Database Management System - Chapter 6
  • 14.
    Image Retrieval Basedon Texture • Texture is described by six features: - coarseness, opposite to fine; - contrast, dynamic range of gray level, ratio of bow areas, sharpness of edges, and period of repeating pattern; - directionality, element shape and placement; - line likeness, shape of a texture element; - regularity, variation of an element placement rule; - roughness.the texture is rough or smooth. Multimedia Database Management System - Chapter 6
  • 15.
    Image Indexing andRetrieval Based on Compressed Image Data • There are three common compression technique for image indexing and retrieval: - DCT Coefficient - Wavelet Coefficient - VQ Compressed Data Multimedia Database Management System - Chapter 6
  • 16.
    Other Techniques •Image Retrieval Based on Model-Based Compression • Image Retrieval Based on Spatial Relationship Multimedia Database Management System - Chapter 6
  • 17.
    Image Retrieval basedon Model-based Compression • An object, is represented by a mathematical model (parameter or mathematical equation). • A very little data required for representing these parameters and equations, so a very high compression can be achieved. • Image distance calculated by parameters differences. Multimedia Database Management System - Chapter 6
  • 18.
    Image Retrieval Basedon Spatial Relationship • A spatial relationship, specifies how some object is located in space in relation to some reference object. • Example queries, “find images containing a sun above to a mountain”. • Example application, Geographical Information System (GIS). Multimedia Database Management System - Chapter 6
  • 19.
    Integrated Image Indexing and Retrieval Techniques Structured attributes Pictorial queries are not supported Integrated IR Techniques + relevance feedback Text-annotation Color based High-level abstractions in images are not supported Shape based Texture based Multimedia Database Management System - Chapter 6