Abstract:The state-of-the-art image classification models, generally including feature coding and pooling, have been widely adopted to generate discriminative and robust image representations. However, the coding schemes available in these models only preserve salient features which results in information loss in the process of generating final image representations. To address this issue, we propose a novel spatial locality-preserving feature coding strategy which selects representative codebook atoms based on their density distribution to retain the structure of features more completely and make representations more descriptive. In the codebook learning stage, we propose an effective approximated K-means with cluster closures to initialize the codebook and independently adjust the center of each cluster of the dense regions. Afterwards, in the coding stage, we first define the concept of “density” to describe the spatial relationship among the code atoms and the features. Then, the responses of local features are adaptively encoded. Finally, in the pooling stage, a locality-preserving pooling strategy is utilized to aggregate the encoded response vectors into a statistical vector for representing the whole image or all the regions of interest. We carry out image classification experiments on three commonly used benchmark datasets including 15-Scene, Caltech-101, and Caltech-256. The experimental results demonstrate that, comparing with the state-of-the-art Bag-of-Words (BoW) based methods, our approach achieves the best classification accuracy on these benchmarked datasets.

Incorporating Spatial Correlogram into Bag-of-features Model for Scene Categorization

Spatial Correlogram Between Codewords: an Effective Representation for Image Classification

Bag of Spatial Visual Words Model for Scene Classification

Incorporating Contextual Information into Bag-of-Visual-Words Framework for Effective Object Categorization

Image Categorization Based On Spatial Visual Vocabulary Model

Sparse code LBP and SIFT features together for scene categorization

Image Classification by Exploiting the Spatial Context Information

Discriminative Spatial Codebook Generation for Image Classification

Spatial graph for image classification

Integrating ILSR to Bag-of-Visual Words Model Based on Sparse Codes of SIFT Features Representations

Spatial Encoding of Visual Words for Image Classification.

Spatially Constrained Sparse Coding Scheme for Natural Scene Categorization

Improving Bag-of-words Scheme for Scene Categorization

Mutual information based codebooks construction for natural scene categorization

Image Classification Based On Local Spatial Pyramid Kernel

Combining LBP and SIFT in Sparse Coding for Categorizing Scene Images.

Image Classification Using Spatial Pyramid Robust Sparse Coding.

Spatial Locality-Preserving Feature Coding for Image Classification

Create visual word pairs dynamically based on sparse codes of SIFT features for image categorization

Spatial Context and Locality-constraint Based Linear Feature Coding

An Integrative Codebook for Natural Scene Categorization