An improved active multi-image super-resolution method via hybrid subpixel images

Multi-image super-resolution (MISR) methods generate a high-resolution image by integrating several low-resolution subpixel images, where various methods focus on improving efficiency by actively acquiring representative subpixel images, with a trade-off in accuracy. We tackled this challenge with an advanced active MISR method, utilizing subpixel images from hybrid sampling rates. By swapping the ideal impulse for a rectangular pulse in our digital image model, our strategy gathers images at targeted sampling rates and locations through subpixel shifts, aiming to build a high-resolution image with minimal input. To test this method, we developed a microscopic imaging setup with a nano-positioning stage, achieving accurate subpixel movement and multiscale sampling. Results showed our method boosted image quality, with an 18.25% rise in Peak Signal-to-Noise Ratio (PSNR) and a 23.51% jump in Structural Similarity Index (SSIM) over current methods, approaching the precision of top-tier techniques. Overall, it improves accuracy while maintaining efficiency. • O ptimized sampling reduces required subpixel images, boosting efficiency and suitability for engineering use. • D eveloped a digital image model using a rectangular pulse to better reflect real sampling and improve reconstruction. • I mproved active MISR with optimized subpixel capture and modeling for higher accuracy and efficiency.