CENTRIST: A Visual Descriptor for Scene Categorization

By Jianxin Wu, James M. Rehg. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2011

781 citations recorded by the source corpus.

CENsus TRansform hISTogram (CENTRIST), a new visual descriptor for recognizing topological places or scene categories, is introduced in this paper. We show that place and scene recognition, especially for indoor environments, require its visual descriptor to possess properties that are different from other vision domains (e.g., object recognition). CENTRIST satisfies these properties and suits the place and scene recognition task. It is a holistic representation and has strong generalizability for category recognition. CENTRIST mainly encodes the structural properties within an image and suppresses detailed textural information. Our experiments demonstrate that CENTRIST outperforms the current state of the art in several place and scene recognition data sets, compared with other descriptors such as SIFT and Gist. Besides, it is easy to implement and evaluates extremely fast.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/435e2a5164c1ee3006c403199d072823d83a5a21

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Exemplar based Deep Discriminative and Shareable Feature Learning for scene image classification (2015, neutral): “There are three types of features: 1) Hand-crafted features, which include SIFT[1], HoG2x2[2, 55, 58], LBP[14], GIST[15], and CENTRIST[56]; 2) The baseline features learned in unsupervised way without encoding discriminative or shareable information, which include feature extracted with Random weigh”
Place categorization and semantic mapping on a mobile robot (2015, methodology): “Wu et al. [11] defined the visual place categorization problem for robotics using a purely appearance-based method based on CENTRIST features [9].”
On Unifying Multi-view Self-Representations for Clustering by Tensor Multi-rank Minimization (2016, positive): “As for scene clustering, some sophisticated features (such as PHOW [42], CENTRIST [41], etc.”
Scene recognition and weakly supervised object localization with deformable part-based models (2011, background): “Table 2 compares our performance with a number of baselines and state-of-the-art approaches [12, 16, 19, 22].”
A Survey on Multi-View Clustering (2017, neutral): “, CENTRIST [133], ColorMoment [134], HOG [135], LBP [136] and SIFT [137] can be extracted from the images (see the Fig.”
Multi-scale Recognition with DAG-CNNs (2015, neutral): “Deep19-DAG 92.9 Deep19 [29] 90.8 Caffe-DAG 89.7 Caffe [14] 86.8 Place [39] 91.6 CENTRIST [36] 84.8 Hybrid [3] 83.7 Spatial pyramid [17] 81.4 Object bank [19] 80.9 Reconﬁgurable model [23] 78.6 Spatial Envelop [22] 74.1 Baseline [9] 65.2 Table 1.”
Local structure preserving discriminative projections for RGB-D sensor-based scene classification (2015, methodology): “By considering the spatial layout of features, spatial pyramid matching (SPM) has been utilized in recent state of the art scene classification systems with some success [5,27,50].”
Towards Scalable Summarization of Consumer Videos Via Sparse Dictionary Selection (2012, methodology): “Popular features are Gist [3] and CENTRIST [4], [5].”
Scene classification based on single-layer SAE and SVM (2015, methodology): “Paper (Wu & Rehg, 2011) emphasize on the indoor scenes, and introduce the visual descriptor, CENsus TRansform hISTogram (CENTRIST), into the scene classification task.”
Remote Sensing Image Classification: No Features, No Clustering (2015, neutral): “census transform histogram (CENTRIST) [12], and local”
Deep learning approaches for COVID-19 detection based on chest X-ray images (2020, positive): “The CENTRIST technique produced accuracy scores of 84.2% for Linear, Quadratic, and Gaussian kernel functions, and 81.1% for the Cubic kernel function.”
Feature Learning for Image Classification Via Multiobjective Genetic Programming (2014, methodology): “For comparison, we have also evaluated some prevalent hand-crafted descriptors, including flattened image, shape context, color histogram, HOG, P-HOG, SIFT, LBP, Gabor bank [24], BIF, HRSE, texton histogram, and Centrist [45].”
Knowledge Guided Disambiguation for Large-Scale Scene Classification With Multi-Resolution CNNs (2016, background): “SCENE recognition [1], [2] is a fundamental problem in computer vision, and has received increasing attention in the past few years [3], [4], [5], [6], [7], [8], [9], [10], [11].”
Multi-view Subspace Clustering (2015, background): “The visual patterns of descriptors CENTRIST, ColorMoment, LBP, HOG and SIFT on three sample images from Caltech101-7.”
Land Use Classification in Remote Sensing Images by Convolutional Neural Networks (2015, methodology): “The interactions among RGB color bands is exploited also in mCENTRIST [27], an extension of the CENTRIST algorithm [28], based in turn on LBP histograms and principal component analysis.”
Deep Learning for Content-Based Image Retrieval: A Comprehensive Study (2014, background): “Over the past decades, a variety of lowlevel feature descriptors have been proposed for image representation [21], ranging from global features, such as color features [21], edge features [21], texture features [32], GIST [36, 37], and CENTRIST [49], and recent local feature representations, such as”
Pairwise Rotation Invariant Co-Occurrence Local Binary Pattern (2014, methodology): “Following the standard protocol [37], [58], we randomly select 25 images per class for training and the rest for testing.”