Parallelization and Optimization of a Method for Web Content Analysis and Accessibility Evaluation for Visually Impaired Users

This paper presents an optimized and parallelized method for automated analysis of text contrast in images in accordance with the WCAG 2.1 accessibility standards. The proposed approach addresses key limitations of the baseline implementation, including low scalability, sequential text block processing, and color detection inaccuracies. To overcome these challenges, several improvements were introduced: image preprocessing (scaling, noise filtering, contrast enhancement), region-of-interest detection, pixel clustering via the K-means algorithm, and parallel processing of key stages using the multiprocessing library. The method was evaluated using a comprehensive test set of images of various sizes and complexity levels. Experimental results demonstrate that the optimized version reduces processing time by up to 60%, improves OCR accuracy by up to 3.4%, and provides higher stability and adaptability to complex image backgrounds compared to existing open-source and commercial solutions.