WORD-LEVEL BLUR FOR REAL-TIME SCREEN ABUSE DETECTION: IMPLEMENTATION AND EVALUATION

computer Engineering

Traditional screen abuse detection systems employ region-based blurring that obscures entire screen sections, causing significant visual disruption and poor user experience. This paper presents a novel word-level blur approach that uses optical character recognition (OCR) combined with natural language processing (NLP) to detect and blur only specific offensive words rather than entire regions. Our system employs a two-stage detection pipeline combining fast wordlist matching with context-aware NLP classification, partitioning the screen into nine parallel processing zones to achieve real-time performance. Experimental evaluation demonstrates 100ms frame processing time with 95%+ accuracy on clear text, representing 78% reduction in visual disruption compared to region-based methods while maintaining equivalent detection rates. The system supports English and Hindi text processing with transliteration handling and operates effectively in both GUI and headless deployment modes. User studies with 50 participants show 91% recommendation rate and significantly improved satisfaction scores over traditional region-based approaches, validating the practical benefits of surgical content filtering over blunt region-based methods.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.