Experiments on Selection of Codebooks for Local Image Feature Histograms

Viitaniemi, Ville; Laaksonen, Jorma

doi:10.1007/978-3-540-85891-1_16

Cited by 20 publications

(20 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We determine the histogram bins by clustering a sample of the interest point SIFT descriptors (20 per image) with the Linde-Buzo-Gray (LBG) algorithm. In our earlier experiments [8] we have found such codebooks to perform reasonably well while the computational cost associated with the clustering still remains manageable. The LBG algorithm produces codebooks with sizes in powers of two.…”

Section: Baseline Systemmentioning

confidence: 86%

See 1 more Smart Citation

Combining Local Feature Histograms of Different Granularities

Viitaniemi¹,

Laaksonen²

2009

Image Analysis

Self Cite

View full text Add to dashboard Cite

Abstract. Histograms of local features have proven to be powerful representations in image category detection. Histograms with different numbers of bins encode the visual information with different granularities. In this paper we experimentally compare techniques for combining different granularities in a way that the resulting descriptors can be used as feature vectors in conventional vector space learning algorithms. In particular, we consider two main approaches: fusing the granularities on SVM kernel level and moving away from binary or hard to soft histograms. We find soft histograms to be a more effective approach, resulting in substantial performance improvement over single-granularity histograms.

show abstract

Section: Baseline Systemmentioning

confidence: 86%

“…Earlier [8] we have performed category detection experiments where we have compared ways to select a codebook for a single histogram representation, with varying histogram sizes. For the experiments we used the images and category detection tasks of the publicly available VOC2007 benchmark.…”

Section: Introductionmentioning

confidence: 99%

Combining Local Feature Histograms of Different Granularities

Viitaniemi¹,

Laaksonen²

2009

Image Analysis

Self Cite

View full text Add to dashboard Cite

show abstract

“…Finally, we emphasize that although the idea of vector quantization by randomly sampled centers was already discussed in [7,24], to the best of our knowledge, this is the first work that presents its statistical consistency analysis.…”

Section: Empirically Estimated Density Function For Q I ðXþmentioning

confidence: 95%

“…Note that since both vector quantization algorithms do not rely on the clustering algorithms to identify visual words, they are in general computationally more efficient. In addition, both algorithms have error bounds decreases at the rate of Oð1= ffiffiffi ffi m p Þ when the number of key points n is large, indicating that they are robust to the number of visual words m. We emphasize that although similar random algorithms for vector quantization have been discussed in [5,17,22,14,24], the purpose of this empirical study is to verify that • simple random algorithms deliver similar performance of object recognition as the clustering based algorithm, and • the random algorithms are robust to the number of visual words, as predicted by the statistical consistency analysis.…”

Section: Empirical Studymentioning

confidence: 96%

“…Unlike the existing statistical approaches for key point quantization that are designed to reduce the training error, the proposed framework generalizes the bag-of-words model by the statistical expectation, making it possible to analyze the statistical consistency of the bag-of-words model. Finally, we would like to point out that although several randomized approaches [12,14,24] have been proposed for key point quantization, none of them provides theoretical analysis on statistical consistency. In contrast, we present not only the theoretic results for the two proposed random algorithms for vector quantization, but also the results of the empirical study with object recognition that support the theoretic claim.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Understanding bag-of-words model: a statistical framework

Zhang

Jin

Zhou

2010

Int. J. Mach. Learn. & Cyber.

1,038

402

View full text Add to dashboard Cite

The bag-of-words model is one of the most popular representation methods for object categorization.The key idea is to quantize each extracted key point into one of visual words, and then represent each image by a histogram of the visual words. For this purpose, a clustering algorithm (e.g., K-means), is generally used for generating the visual words. Although a number of studies have shown encouraging results of the bag-of-words representation for object categorization, theoretical studies on properties of the bag-of-words model is almost untouched, possibly due to the difficulty introduced by using a heuristic clustering process. In this paper, we present a statistical framework which generalizes the bag-of-words representation. In this framework, the visual words are generated by a statistical process rather than using a clustering algorithm, while the empirical performance is competitive to clustering-based method. A theoretical analysis based on statistical consistency is presented for the proposed framework. Moreover, based on the framework we developed two algorithms which do not rely on clustering, while achieving competitive performance in object categorization when compared to clustering-based bag-ofwords representations.

show abstract

Open Issues on Codebook Generation in Image Classification Tasks

Piras

Giacinto

2014

Machine Learning and Data Mining in Pattern Recognition

View full text Add to dashboard Cite

Experiments on Selection of Codebooks for Local Image Feature Histograms

Cited by 20 publications

References 7 publications

Combining Local Feature Histograms of Different Granularities

Combining Local Feature Histograms of Different Granularities

Understanding bag-of-words model: a statistical framework

Open Issues on Codebook Generation in Image Classification Tasks

Contact Info

Product

Resources

About