Edge-Preserving Multi-Scale Texture Enhancement and Intelligent Feature Selection for High-Accuracy Image Classification

Abstract

Accurate image classification remains a critical challenge in computer vision due to variations in illumination, noise, and complex texture distributions. This paper presents an Edge-Preserving Multi-Scale Texture Enhancement with Intelligent Feature Selection (EMTE-IFS) framework to improve classification performance. Initially, a multi-scale texture enhancement mechanism is applied to extract both fine and coarse-grained features from input images. An edge-preserving filtering technique is incorporated to retain important structural boundaries while suppressing noise and irrelevant variations, thereby improving feature quality. Subsequently, an intelligent feature selection strategy is employed to identify the most discriminative and non-redundant features from the enhanced feature space. This process reduces dimensionality and computational overhead while maintaining high representational efficiency. The optimized feature subset is then utilized by a robust classification model to achieve superior predictive performance. Experimental results on benchmark datasets demonstrate that the proposed EMTE-IFS framework achieves significant improvements in classification accuracy, precision, recall, and F1-score compared to existing methods. Furthermore, the model exhibits strong robustness under noisy and varying environmental conditions. The proposed approach is well-suited for real-world applications including medical image analysis, remote sensing, and object recognition systems.