Local Regional Object based Retrieval Techniques for CBIR Applications
- Issued Date
- 2008
- Abstract
- The study mainly covers the subject of content based image retrieval (CBIR). Some specialized applications for image based query and retrieval are researched and some useful results are produced. The research includes image structural and color feature analysis in order to ease retrieval of similar images from huge databases. HSV, RGB and gray-scale nature of images was analysed and most prominent and noteworthy parameters were short listed and ranked against various sorts of images.
Many references have been studied in order to rank various notable features of images, which prove key indices or identification marks of various images. Under the light of most reputable and most remarkable research, the features were rigorously tested and confirmed for their solidity. The most useful of them were used to devise several algorithms that can achieve promising retrieval results. Simulation was performed using MATLAB, and tested against existing algorithms.
Several valuable image characteristics were studied separately and collectively in the form of sets. The most useful combinations are devised after collective and iterative simulation of the most fruitful set of image identification features.
- Alternative Title
- Local Regional Object based Retrieval Techniques for CBIR Applications
- Alternative Author(s)
- Waqas Rasheed
- Affiliation
- 일반대학원 정보통신공학과
- Department
- 일반대학원 정보통신공학과
- Advisor
- 박종안
- Awarded Date
- 2009-02
- Table Of Contents
- List of Figures iv
List of Tables vi
A b s t r a c t vii
I. Introduction 1
A. History of CBIR 1
B. Research Background 1
II. Image Retrieval Algorithm based on Incremental CBIR using Color Histogram 4
A. Introduction Picture Expert Group) 4
B. Related Work 4
C. Feature Extraction Algorithm 5
D. Incremental Image Retrieval Approach 6
1. First Level Retrieval 6
2. Second Level Retrieval 7
D. Results and Discussion 8
E. Conclusion 8
III. Image Retrieval using Maximum Frequency of Local Histogram based Color Correlogram 9
A. Introduction 9
B. Method 10
1. Pre-processing 11
2. Splitting Histogram Values by Fixed Frequency Range 11
3. Sub-divisions of Histogram Values from each Bin 12
4. Calculating maximum Frequency of the most Recurring Value From the Bin Sub-divisions 12
5. Correlogram Formation 13
6. Similarity Measurement 13
C. Results 14
D. Conclusion & Discussion 17
IV. Key Objects based Profile for a Content-based Video Information Retrieval and Streaming System using Viewpoint Invariant Regions 18
A. Introduction 18
B. Information Acquisition Process 19
1. Object Recognition 19
2. Object Logging 22
C. Streaming and Support 23
D. Results 24
E. Conclusion & Discussion 25
V. Defining a new Feature Set for Content-based Image Analysis using Histogram Refinement 27
A. Introduction 27
B. Related work 29
C. Proposed Algorithm and Feature Selection 32
1. Pre-processing 32
2. Features from Quantized Bins 33
3. Features from Coherent Clusters 34
4. Additional Features based on Size of Cluster 35
D. Image Retrieval 36
1. Stage 1 36
2. Stage 2 37
3. Stage 3 37
E. Results 38
F. Conclusion & Discussion 44
VI. Petite Feature Set Defining Solid Structural and Color Assets in CBIR Processes 46
A. Introduction 46
B. Related work 47
C. Method 49
1. Processing of individual RGB features of an image 50
2. Processing of overall appearance features of an image 52
D. Results 53
E. Conclusion & Discussion 55
VII. Sum of Values of Local Histograms for Image Retrieval 57
A. Introduction 57
B. Method 58
1. Pre-processing 58
2. Splitting Histogram Values by Fixed Frequency Range 59
3. Calculating sum of values from the bin sub-divisions 60
4. Storing information 60
5. Similarity Measurement 61
C. Results 61
D. Conclusion & Discussion 64
References 65
- Degree
- Master
- Publisher
- 조선대학교
- Citation
- 라쉬드 와카스. (2008). Local Regional Object based Retrieval Techniques for CBIR Applications.
- Type
- Dissertation
- URI
- https://oak.chosun.ac.kr/handle/2020.oak/7411
http://chosun.dcollection.net/common/orgView/200000237390
- Appears in Collections:
- General Graduate School > 3. Theses(Master)
- Files in This Item:
-
-
Download
Local Regional Object based Retrieval Techniques for CBIR Applications.pdf
기타 데이터 / 2 MB / Adobe PDF
-
Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.