Because the pixel distribution in detail-deficient images is disrupted, direct enhancement methods often fail to produce satisfactory results. Therefore, an intelligent enhancement processing method of visual image details based on data decomposition and the Retinex algorithm is proposed. To address irregular pixel distributions, we apply Arnold transform to scramble the original image with blurred details. During cyclic iteration, each transformation’s output becomes the next input, creating a continuous process, and Logistic mapping decomposes the image; In the enhancement phase, we introduce a multi-scale Retinex (MSR) algorithm. Using Gaussian filters with varying scales, it estimates illumination for low-light images. We then compensate illumination and incorporate weights into MSR. Tests show that for various blurred images, our method achieves PSNR > 19.0 and SSIM > 0.90, outperforming comparative methods from literature [7, 8].