Abstract
Hyperspectral image (HSI) classification, which aims to assign an accurate label for hyperspectral pixels, has drawn great interest in recent years. Although low rank representation (LRR) has been used to classify HSI, its ability to segment each class from the whole HSI data has not been exploited fully yet. LRR has a good capacity to capture the underlying low-dimensional subspaces embedded in original data. However, there are still two drawbacks for LRR. First, LRR does not consider the local geometric structure within data, which makes the local correlation among neighboring data easily ignored. Second, the representation obtained by solving LRR is not discriminative enough to separate different data. In this paper, a novel locality and structure regularized low rank representation (LSLRR) model is proposed for HSI classification. To overcome the above limitations, we present locality constraint criterion (LCC) and structure preserving strategy (SPS) to improve the classical LRR. Specifically, we introduce a new distance metric, which combines both spatial and spectral features, to explore the local similarity of pixels. Thus, the global and local structures of HSI data can be exploited sufficiently. Besides, we propose a structure constraint to make the representation have a near block-diagonal structure. This helps to determine the final classification labels directly. Extensive experiments have been conducted on three popular HSI datasets. And the experimental results demonstrate that the proposed LSLRR outperforms other state-of-the-art methods.
Introduction
Hyperspectral images (HSIs) capture hundreds of continuous spectral bands per pixel, providing rich information about land materials. However, classifying these pixels is notoriously difficult due to the "Hughes phenomenon" (where high dimensionality degrades classifier performance) and a chronic lack of labeled training data.
To solve this, researchers often use Low Rank Representation (LRR), a technique that assumes high-dimensional data actually lies in multiple low-dimensional subspaces. While LRR is great at capturing the global structure of data, standard LRR has two major blind spots: it ignores local spatial context by treating pixels as independent, and it lacks a discriminative structure. Ideally, a representation matrix should be "block-diagonal" (where samples of Class A only correlate with Class A). Standard LRR struggles to form this clean separation, making the final classification step harder.
Proposed Method: LSLRR
The proposed Locality and Structure Regularized Low Rank Representation (LSLRR) model upgrades standard LRR with three key innovations to address these limitations.
Locality Constraint Criterion (LCC)
The authors introduce a new distance metric that fuses spectral features (the pixel's actual data) with spatial features (the pixel's physical coordinates). A hyperparameter ($m$) balances the two. This metric acts as a penalty term during optimization, forcing the model to respect the local geometric structure of the image. Relying solely on spectral signatures often leads to noisy, "salt-and-pepper" classification maps; forcing the model to acknowledge physical distance drastically smooths and corrects predictions.
Structure Preserving Strategy (SPS)
To fix the "messy matrix" problem, the authors introduce an auxiliary matrix that forces the learned representation into a class-wise block-diagonal structure. By utilizing the labels of the training data and a Gaussian similarity function for the testing data, the model strictly groups similar classes together. This block-diagonal structure eliminates the need for complex downstream classifiers; the model simply looks at which block a test pixel's representation falls into and assigns the label via basic summation.
Dictionary Learning
Standard LRR uses all data samples as its "dictionary" to reconstruct the image, which is computationally heavy and introduces noise. LSLRR instead iteratively learns a compact, discriminative dictionary from the data, filtering out corrupted samples. The resulting objective function is solved using the high-efficiency Inexact Augmented Lagrange Multiplier (IALM) algorithm.
Experimental Results
The model was evaluated on three standard benchmarks: Indian Pines (agricultural), Pavia University (urban), and Salinas (agricultural).
LSLRR demonstrated massive improvements over the baseline LRR model. On the Indian Pines dataset, LSLRR achieved an Overall Accuracy (OA) of 95.63%, compared to standard LRR's 70.47%—a jump of over 25%. Across all datasets, LSLRR outperformed advanced competitors like SVM with Composite Kernels (SVMCK) and Joint Robust Sparse Representation (JRSRC), achieving 98.52% OA on Pavia University and 97.77% OA on Salinas. Notably, the model maintained superior accuracy even when training data was restricted to just 5% or 10% of total labeled pixels.
Conclusion
LSLRR effectively combines global low-rank structures with local spatial-spectral constraints to produce highly discriminative classification maps. While it achieves state-of-the-art accuracy, its computational complexity is higher than simple pixel-wise classifiers, and it requires careful manual tuning of hyperparameters (like the spatial-spectral balance $m$) for different geographic scenes. Future work will focus on automated or adaptive parameter selection mechanisms to improve usability across diverse datasets.
Summary updated