Blockwise Statistical Analysis and Processing of Large Images

Abstract

This study introduces a novel and efficient method for large-scale image processing by integrating statistical analysis with a blockwise framework. Our approach addresses the fundamental memory and computational limitations of traditional image processing techniques by dividing the image into smaller blocks, thereby enabling the analysis of datasets that exceed available memory. Our method's core innovation lies in a two-phase algorithmic design that efficiently computes global image statistics and consistently applies them to each block. To validate this approach, we developed a MATLAB-based model and performed a comprehensive quantitative and visual comparison against traditional and state-of-the-art alternatives. The experimental results demonstrated the superior performance of our proposed method, achieving a peak signal-to-noise ratio (PSNR) of 24.248 and a structural similarity index (SSIM) of 0.754. These metrics significantly surpassed those of a hierarchical pyramid method (PSNR: 18.434, SSIM: 0.539) and a simple local tiling approach (PSNR: 12.415, SSIM: 0.681). Furthermore, our method proved to be exceptionally efficient, with a low execution time and minimal memory usage, validating its scalability for large-scale datasets. This study provides valuable insights into the application of blockwise statistical analysis, offering a robust and practical solution for researchers and professionals in fields such as remote sensing, medical imaging, and computer vision. Our findings contribute to the advancement of image processing methods for handling the ever-growing size of modern image collections.