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.

Feature Normalization For Part-Based Image Classification

Simple Techniques Make Sense: Feature Pooling and Normalization for Image Classification

Normalization in Proportional Feature Spaces

Does Normalization Methods Play a Role for Hyperspectral Image Classification?

Image Classification Method Rationally Utilizing Spatial Information of the Image

SS-Norm: Spectral-spatial Normalization for Single-Domain Generalization with Application to Retinal Vessel Segmentation.

Learning Relative Features Through Adaptive Pooling for Image Classification

Normalization of Face Illumination Based on Large-and Small-Scale Features

Unified Batch Normalization: Identifying and Alleviating the Feature Condensation in Batch Normalization and a Unified Framework

Considering The Spatial Layout Information Of Bag Of Features (Bof) Framework For Image Classification

Spatial pooling of heterogeneous features for image classification.

Bilaterally Normalized Scale-Consistent Sinkhorn Distance for Few-Shot Image Classification

Image Classification Based on Nearest Neighbor Basis Vectors

Generalized regular spatial pooling for image classification

Iris Normalization Beyond Appr-Circular Parameter Estimation

PointNorm: Dual Normalization is All You Need for Point Cloud Analysis

Spatial Locality-Preserving Feature Coding for Image Classification

BCN: Batch Channel Normalization for Image Classification

Kernelized Pyramid Nearest-Neighbor Search for Object Categorization

Bag of Surrogate Parts Feature for Visual Recognition

Geometric L(P)-Norm Feature Pooling for Image Classification