This study investigates the impact of advanced image processing techniques on image captioning performance using the COCO dataset. We applied six techniques—Unsharp Masking, High Pass Filter, Contrast Enhancement, Laplacian Sharpening, Guided Image Filtering, and Total Variation Minimization—to LL band-filtered images. The processed images were evaluated using Peak Signal-to-Noise Ratio (PSNR), Mean Squared Error (MSE), and Structural Similarity Index (SSIM) are utilized for the image quality. Additionally, we utilized the Bootstrapping Language-Image Pre-training (BLIP) framework to generate captions and analyzed the results using BLEU, METEOR, ROUGE_L, CIDEr, and SPICE metrics. Our results demonstrate that Guided Image Filtering and Total Variation Minimization significantly enhance image quality and caption accuracy, with Guided Image Filtering achieving the highest improvements across most metrics. This study highlights the importance of image processing in refining visual inputs for caption generation, offering valuable insights into the optimization of image quality for machine learning applications.