To address the multimodal transformation problem between on-site dust footprints and color pressure footprints, an improved Pix2Pix network-based footprint transformation algorithm is proposed. First, an attention module is incorporated into the generator’s skip connections to suppress interference from background noise and illumination noise in dust footprints, while residual connections are added to the innermost layer of the generator to enhance the transmission of key footprint features. Second, a weighted fusion discriminator is constructed by combining the dual discrimination mechanisms of PatchGAN and the PixelGAN, enabling local texture discrimination and pixel-level consistency discrimination between generated footprint images and the target color pressure footprint images. Then, a multi-task weighted loss function is designed, comprising Least Squares GAN Loss (LSGAN), Perceptual Loss (Perc), and L1 Loss, with the adversarial loss weight λ_GAN=1, perceptual loss weight λ_P=10, and L1 loss weight λ_L1=100. Finally, the improved Pix2Pix network is evaluated qualitatively and quantitatively on both training and test datasets. Experimental results show that compared with the baseline network, the improved Pix2Pix network generates footprint images with more complete contours, clearer textures, and stronger visual consistency. The Structural Similarity Index Measure (SSIM), Cosine Similarity (CosSim), and Peak Signal-to-Noise Ratio (PSNR) metrics are improved by 12.7%, 19.3% and 10.8%, respectively, demonstrating its effectiveness in realizing the efficient mutual transformation between dust footprints and color pressure footprints.