Optimizing Feature Extraction with Multi-Variant Bidirectional PCA for 2D Image Analysis

Effective feature extraction is essential for 2D image analysis, particularly in handling high-dimensional data, local pattern variations, and lighting changes. This study introduces Global-Local Bidirectional PCA (GL-BiPCA), a new method that integrates global and local features in both row and column directions using multi-variant PCA. Tested on a self-acquired dataset of pineapple images captured at different ripeness stages under controlled lighting, GL-BiPCA achieved a classification accuracy of 99.67%. The method's multi-level dimensionality reduction significantly reduced space requirements and resulted in fast recognition times. These results demonstrate that GL-BiPCA optimizes both performance and efficiency, making it a valuable tool for accurate and computationally efficient image recognition. Future research should explore its application across diverse datasets and real-world scenarios to further validate its effectiveness and efficiency.