图像处理方法中图像增强传统方法的研究综述

A Review of Traditional Image Enhancement Methods in Image Processing

摘要: 目的：图像增强是图像处理中的关键技术，旨在改善图像的视觉质量或突出特定信息，以满足后续分析、识别或显示的需求。其核心目标是提升对比度、亮度及细节表现，同时抑制噪声，适用于医学影像、遥感、安防监控等多个领域。方法：图像增强方法主要分为空域和频域两大类。空域方法直接操作像素值，包括：直方图均衡化(全局/CLAHE)通过重新分布像素灰度增强对比度；线性变换与Gamma校正调整亮度和对比度；Retinex算法分离光照与反射分量以改善光照不均。频域方法(如傅里叶变换、小波变换)通过滤波增强特定频率成分。近年来，深度学习方法(如生成对抗网络)通过数据驱动实现自适应增强，显著提升了复杂场景下的性能。传统方法计算高效但依赖参数调优，深度学习方法灵活性高但需大量数据。未来研究将聚焦于结合传统方法与深度学习，以平衡实时性、鲁棒性和泛化能力。

Abstract: Objective: Image enhancement is a key technology in image processing, which aims to improve the visual quality of images or highlight specific information to meet the needs of subsequent analysis, recognition or display. Its core goal is to improve contrast, brightness and detail performance, while suppressing noise, which is suitable for medical imaging, remote sensing, security monitoring and other fields. Methods: Image enhancement methods are mainly divided into two categories: spatial domain and frequency domain. Spatial methods manipulate pixel values directly, including: histogram equalization (global/CLAHE) to enhance contrast by redistributing pixel grayscale; Linear transformation with gamma correction to adjust brightness and contrast; The Retinex algorithm separates the light and reflection components to improve uneven lighting. Frequency-domain methods (e.g., Fourier transform, wavelet transform) enhance specific frequency components through filtering. In recent years, deep learning methods, such as generative adversarial networks, have achieved adaptive enhancement through data-driven, which has significantly improved the performance in complex scenarios. The traditional method is efficient but relies on parameter tuning, while the deep learning method is flexible but requires a large amount of data. Future research will focus on combining traditional methods with deep learning to balance real-time, robustness, and generalization capabilities.