UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)
UvA-DARE (Digital Academic Repository)
Invariant color descriptors for efficient object recognition
van de Sande, K.E.A.
Publication date 2011
Link to publication
Citation for published version (APA):
van de Sande, K. E. A. (2011). Invariant color descriptors for efficient object recognition.
General rights
It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).
Disclaimer/Complaints regulations
If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.
Contents
1 Introduction 1
1.1 Object appearance in the world . . . 2
1.2 ‘What’ . . . 3
1.3 ‘Where’ . . . 5
1.4 Organization of the Thesis . . . 5
2 Evaluating Color Descriptors for Object and Scene Recognition 7 2.1 Introduction . . . 7
2.2 Reflectance Model . . . 8
2.2.1 Diagonal Model . . . 9
2.2.2 Photometric Analysis . . . 10
2.3 Color Descriptors and Invariant Properties . . . 11
2.3.1 Histograms . . . 11
2.3.2 Color Moments and Moment Invariants . . . 13
2.3.3 Color SIFT Descriptors . . . 14
2.3.4 Conclusion . . . 15
2.4 Experimental Setup . . . 15
2.4.1 Feature Extraction Pipelines . . . 16
2.4.2 Classification . . . 17
2.4.3 Experiment 1: Illumination Changes . . . 19
2.4.4 Experiment 2: Image Benchmark . . . 19
2.4.5 Experiment 3: Video Benchmark . . . 19
2.4.6 Evaluation Criteria . . . 20
2.5 Results . . . 20
2.5.1 Experiment 1: Illumination Changes . . . 20
2.5.2 Experiment 2: Image Benchmark . . . 24
2.5.3 Experiment 3: Video Benchmark . . . 26
2.5.4 Comparison with state-of-the-art . . . 26
ii CONTENTS
2.5.5 Discussion . . . 30
2.6 Conclusion . . . 32
3 Illumination-Invariant Descriptors for Discriminative Visual Object Cate-gorization 33 3.1 Introduction . . . 33
3.2 Illumination-Invariant Descriptors . . . 34
3.2.1 Introduction . . . 34
3.2.2 Diagonal Model . . . 34
3.2.3 A Novel Class of Illumination-Invariant SIFT Descriptors . . . 35
3.2.4 Instantiating Illumination-Invariant Descriptors . . . 38
3.2.5 Discussion . . . 39
3.3 Methods and Experimental Setup . . . 39
3.3.1 Feature Extraction . . . 39
3.3.2 Category Model Training . . . 40
3.3.3 Experiment 1: Candidate Descriptor Selection . . . 41
3.3.4 Experiment 2: Multiple Kernel Learning . . . 41
3.3.5 Experiment 3: Optimized Multi-Channel Descriptors . . . 41
3.3.6 Datasets . . . 42
3.3.7 Evaluation criteria . . . 42
3.4 Results . . . 43
3.4.1 Experiment 1: Candidate Descriptor Selection . . . 43
3.4.2 Experiment 2: Multiple Kernel Learning . . . 43
3.4.3 Experiment 3: Optimized Multi-Channel Descriptors . . . 45
3.4.4 Application: Object Localisation . . . 48
3.5 Conclusion . . . 49
4 Empowering Visual Categorization with the GPU 51 4.1 Introduction . . . 51
4.2 Overview of Visual Categorization . . . 53
4.2.1 Image Feature Extraction . . . 53
4.2.2 Category Model Learning . . . 57
4.2.3 Test Image Classification . . . 58
4.3 GPU Accelerated Categorization . . . 58
4.3.1 Parallel Programming on the GPU and CPU . . . 59
4.3.2 Algorithm 1: GPU-Accelerated Vector Quantization . . . 59
4.3.3 Algorithm 2: GPU-Accelerated Kernel Value Precomputation . . . 61
4.4 Experimental Setup . . . 62
4.4.1 Experiment 1: Vector Quantization Speed . . . 63
4.4.2 Experiment 2: Kernel Value Precomputation Speed . . . 63
4.4.3 Experiment 3: Visual Categorization Throughput . . . 63
4.5 Results . . . 64
Contents iii
4.5.2 Experiment 2: Kernel Value Precomputation Speed . . . 65
4.5.3 Experiment 3: Visual Categorization Throughput . . . 65
4.6 Other Applications . . . 67
4.6.1 Application 1: k-means Clustering . . . 67
4.6.2 Application 2: Bag-of-Words Model for Text Retrieval . . . 67
4.6.3 Application 3: Multi-Frame Processing for Video Retrieval . . . 68
4.7 Conclusions . . . 68
5 Segmentation as Selective Search for Object Recognition 71 5.1 Introduction . . . 71
5.2 Related Work . . . 73
5.2.1 Exhaustive Search for Recognition . . . 73
5.2.2 Selective Search for Object Delineation . . . 75
5.3 Segmentation as Selective Search . . . 75
5.3.1 Our Segmentation Algorithm . . . 76
5.3.2 Shadow, Shading and Highlight Edges . . . 77
5.3.3 Discussion . . . 77
5.4 Object Recognition System . . . 77
5.5 Evaluation . . . 79
5.5.1 Exp. 1: Segmentation for Selective Search . . . 80
5.5.2 Exp. 2: Selective Search for Recognition . . . 81
5.5.3 Exp. 3: Selective Search for Object Delineation . . . 83
5.5.4 Exp. 4: Object Recognition Accuracy . . . 85
5.6 Conclusions . . . 87
6 Summary and Conclusions 89 6.1 Summary . . . 89 6.2 Conclusions . . . 91 Bibliography 102 Samenvatting 103 Dankwoord 107 Biography 109