The enhanced LBG algorithm

Patanè, Salvatore; Russo, Marco

doi:10.1016/s0893-6080(01)00104-6

Cited by 196 publications

(119 citation statements)

References 21 publications

Supporting

Mentioning

110

Contrasting

Unclassified

Order By: Relevance

“…One can see that results are not regular according to dictionary sizes. This is due to the learning of dictionary: even if we use a K-Means algorithm which gives almost stable dictionaries [31], they are still subject to some variations. However, one can see that there is emerging combinations.…”

Section: Global Dictionarymentioning

confidence: 99%

Incremental kernel learning for active image retrieval without global dictionaries

Gosselin

Precioso

Philipp‐Foliguet

2011

Pattern Recognition

View full text Add to dashboard Cite

In content-based image retrieval context, a classic strategy consists in computing off-line a dictionary of visual features. This visual dictionary is then used to provide a new representation of the data which should ease any task of classification or retrieval. This strategy, based on past research works in text retrieval, is suitable for the context of batch learning, when a large training set can be built either by using a strong prior knowledge of data semantics (like for textual data) or with an expensive off-line pre-computation. Such an approach has major drawbacks in the context of interactive retrieval, where the user iteratively builds the training data set in a semi-supervised approach by providing positive and negative annotations to the system in the relevance feedback loop. The training set is thus built for each retrieval session without any prior knowledge about the concepts of interest for this session. We propose a completely different approach to build the dictionary on-line from features extracted in relevant images. We design the corresponding kernel function, which is learnt during the retrieval session. For each new label, the kernel function is updated with a complexity linear with respect to the size of the database. We propose an efficient active learning strategy for the weakly supervised retrieval method developed in this paper. Moreover this framework allows the combination of features of different types. Experiments are carried out on standard databases, and show that a small dictionary can be dynamically extracted from the features with better performances than a global one.

show abstract

Section: Global Dictionarymentioning

confidence: 99%

Incremental kernel learning for active image retrieval without global dictionaries

Gosselin

Precioso

Philipp‐Foliguet

2011

Pattern Recognition

View full text Add to dashboard Cite

show abstract

“…When each pixel of an MS image is mapped onto a color space partitioned into a set of mutually exclusive and totally exhaustive hyperpolyhedra equivalent to a vocabulary of BC names, then a 2D multilevel color map (2D gridded dataset of a multilevel variable) is generated automatically (without human-machine interaction) in near real-time (with computational complexity increasing linearly with image size), where the number k of 2D map levels (color strata, color names) belongs to range {1, ColorVocabularyCardinality}. Popular synonyms of measurement space hyperpolyhedralization (discretization, partition) are vector quantization (VQ) in inductive machine learning-from-data (Cherkassky & Mulier, 1998; Elkan, 2003; Fritzke, 1997a, 1997b; Lee, Baek, & Sung, 1997; Linde, Buzo, & Gray, 1980; Lloyd, 1982; Patanè and Russo, 2001, 2002), and deductive fuzzification of a numeric variable into fuzzy sets in fuzzy logic (Zadeh, 1965). Typical inductive learning-from-data VQ algorithms aim at minimizing a known VQ error function, e.g., a root mean square vector quantization error (RMSE), given a number of k discretization levels selected by a user based on a priori knowledge and/or heuristic criteria.…”

Section: Introductionmentioning

confidence: 99%

“…For example, in a bag-of-words model applied to CV tasks, a numeric color space is typically discretized into a categorical color variable (codebook of codewords) by an inductive VQ algorithm, such as k -means; next, the categorical color variable is simplified by a 1st-order histogram representation, which disregards word grammar, semantics and even word-order, but keeps multiplicity; finally, the frequency of each color codeword is used as a feature for training a supervised data learning classifier (Cimpoi et al, 2014). Unlike the k -means VQ algorithm where the system’s free-parameter k is user-defined based on heuristics and the VQ error is estimated from the unlabeled dataset at hand, a user can fix the target VQ error value, so that it is the free-parameter k to be dynamically learned from the finite unlabeled dataset at hand by an inductive VQ algorithm (Patané & Russo, 2001, 2002), such as ISODATA (Memarsadeghi, Mount, Netanyahu, & Le Moigne, 2007). It means there is no universal number k of static hyperpolyhedra in a vector data space suitable for satisfying any VQ problem of interest if no target VQ error is specified in advance.…”

Section: Introductionmentioning

confidence: 99%

GEO-CEOS stage 4 validation of the Satellite Image Automatic Mapper lightweight computer program for ESA Earth observation level 2 product generation - Part 1: Theory

et al. 2018

View full text Add to dashboard Cite

ESA defines as Earth Observation (EO) Level 2 information product a single-date multi-spectral (MS) image corrected for atmospheric, adjacency and topographic effects, stacked with its data-derived scene classification map (SCM), whose legend includes quality layers cloud and cloud-shadow. No ESA EO Level 2 product has ever been systematically generated at the ground segment. To fill the information gap from EO big data to ESA EO Level 2 product in compliance with the GEO-CEOS stage 4 validation (Val) guidelines, an off-the-shelf Satellite Image Automatic Mapper (SIAM) lightweight computer program was validated by independent means on an annual 30 m resolution Web-Enabled Landsat Data (WELD) image composite time-series of the conterminous U.S. (CONUS) for the years 2006–2009. The SIAM core is a prior knowledge-based decision tree for MS reflectance space hyperpolyhedralization into static color names. Typically, a vocabulary of MS color names in a MS data (hyper)cube and a dictionary of land cover (LC) class names in the scene-domain do not coincide and must be harmonized (reconciled). The present Part 1—Theory provides the multidisciplinary background of a priori color naming. The subsequent Part 2—Validation accomplishes a GEO-CEOS stage 4 Val of the test SIAM-WELD annual map time-series in comparison with a reference 30 m resolution 16-class USGS National Land Cover Data 2006 map, based on an original protocol for wall-to-wall thematic map quality assessment without sampling, where the test and reference maps feature the same spatial resolution and spatial extent, but whose legends differ and must be harmonized.

show abstract

“…In VQ, the most important task is designing an efficient codebook. There are already several algorithms [6][7][8][9][10][11][12] published on how to generate a codebook. The LBG algorithm [6] is the most cited and widely used algorithm on designing the VQ codebook.…”

Section: Introductionmentioning

confidence: 99%

“…Due to the bad codebook initialization, it always converges to the nearest local minimum. This problem is called the local optimal problem [10]. In addition, it is observed that the time required to complete the iterations depends upon how good the initial codebook is.…”

Section: Introductionmentioning

confidence: 99%