Enhanced retinex image processing via gradient domain guided image filtering and adaptive NLM

To address the challenges of low-light (LOL), non-uniform illumination, and underwater images, we propose an enhanced Retinex image processing method with strong generalization capability. The research framework consists of the following key components: First, the raw RGB image is converted to Luminance-Inphase-Quadrature (YIQ) color space to separate luminance and chrominance components. Subsequently, the luminance component is enhanced using a Multi-Scale Retinex (MSR) algorithm based on Gradient Domain Guided Image Filtering (GDGIF). At the same time, adaptive Non-Local Means (NLM) filtering is introduced for noise suppression. For chrominance enhancement, a nonlinear gamma model is constructed to expand the display dynamic range. Finally, the processed YIQ image is reconstructed into RGB format. Experimental results on LOL, non-uniform illumination, and underwater image datasets demonstrate that the proposed method significantly outperforms traditional enhancement algorithms in brightness adjustment, structure preservation, and detail clarity, providing an effective solution for image quality improvement in the complex image enhancement domain.

کلیدواژه ها:

Multi Scale Retinex ، Gradient Domain Guided Image Filter ، Adaptive Non-Local Means ، gamma