Bringing Lost Scripts Into Focus

“Stone inscriptions carry centuries of human stories, yet their degraded conditions make automatic processing extremely challenging. Our patching strategy not only improves binarization, it helps models ‘see’ inscriptions the way a human expert would — focusing on meaningful shapes in context.”— Dr. Ravi Kiran Sarvadevabhatla, Principal Investigator.

Stone inscriptions are invaluable reservoirs of history, language, and culture — but they pose massive challenges for digital analysis. Weathering, low contrast between etched text and stone surface, and irregular layouts make traditional image processing and OCR techniques struggle. The team’s work tackles this head-on with a novel binarization strategy that significantly improves how text is separated from complex backgrounds.

“We wanted to build something that goes beyond traditional thresholds and handcrafted filters — something that can handle real-world inscription conditions. This award motivates us to push further.”— Amal Joseph, Co-author

Their approach introduces a character-context-aware patching strategy to train deep learning models that can focus on fine textual structures while understanding the context around them. This enables robust binarization even when the inscriptions are faint or obscured by noise, opening doors to further processing like script identification and automatic recognition.

This research shines for several reasons:

• Robustness in Real-World Conditions — Unlike many existing algorithms that perform well only in controlled settings, this method can handle real and degraded inscriptions.

• Multidisciplinary Impact — The outcomes are valuable not only for computer scientists but also for archaeologists, historians, and epigraphists who work to preserve and understand historical texts.

• Recognition at a National Stage — Competing with numerous high-quality submissions from across India, the team’s work stood out for its creativity, technical depth, and practical relevance, earning the Best Paper Runner-Up Award at one of the nation’s foremost conferences in vision and graphics.

“Winning this recognition at ICVGIP was a proud moment for us. Presenting the poster and engaging with peers has given us fresh perspectives for the next stage of this research.”

— Pratyush Jena, Co-author

Celebrating Young Innovators

While Dr. Ravi Kiran provided visionary guidance, the contributions of his students — Arnav Sharma, Amal Joseph, and Prayush Jena — were instrumental in turning the research into an award-winning success. Their passion for blending classical problems with modern computing speaks to a growing generation of innovators pushing boundaries at the intersection of technology and heritage.

“Our strategy adapts to different scripts and textures, showing promising generalization. This project blends cultural heritage and deep learning in a way that truly inspires us.”

— Arnav Sharma, Co-author

A big round of applause to Dr. Ravi Kiran Sarvadevabhatla, Pratyush Jena, Amal Joseph, and Arnav Sharma for this outstanding contribution to the field of vision and image processing — and for helping unlock the secrets etched in stone.

Project page: : https://ihdia.iiit.ac.in/shilalekhya-binarization