Delhi Government's Duplicate Image Purge: What Changed This Week in the City's Digital Records Drive

Delhi's Department of Information Technology flagged more than 1.2 lakh duplicate images across government-held databases this week, triggering an urgent cleanup exercise that administrators say is the largest of its kind the capital has attempted. The exercise, centred on voter identity photographs and civic permit records, has been under discussion since early 2025 but only moved into execution this fortnight after the Election Commission of India set a hard deadline for clean electoral rolls ahead of the next scheduled municipal cycle.

The timing matters for reasons that go beyond bureaucratic tidiness. Duplicate images in voter rolls have been a live political controversy in Delhi for years, with the Aam Aadmi Party and the Bharatiya Janata Party trading accusations over inflated or suppressed voter lists in constituencies spanning Chandni Chowk, Okhla, and Rohini. A clean image database directly affects whether a voter's entry survives de-duplication algorithms — and in a city where individual assembly segments can be decided by margins under two thousand votes, the stakes are concrete.

Where the Problem Is Worst — and Why

The duplication problem is not evenly distributed. Officials at the Delhi State Election Commission office on Patel Chowk have identified three broad categories: scanned photographs uploaded multiple times during digitisation drives in 2018 and 2021, images with identical pixel data but different metadata entries, and low-resolution photographs that matching software struggles to distinguish from one another. Old Delhi localities including Matia Mahal and Ballimaran have particularly dense duplication rates, partly because paper rolls from those wards were digitised by multiple contractors over separate drives with no unified deduplication check applied at the time.

The Delhi Metro Rail Corporation is separately grappling with a parallel problem in its Phase 4 construction documentation. Permit applications for stations along the Janakpuri West–Krishna Park Extension corridor and the Aerocity–Tughlakabad line submitted hundreds of site photographs between 2022 and 2024. An internal audit, the findings of which were circulated to contractors in June 2026, found that roughly 18 percent of those images were either duplicates or near-duplicates that had been filed under different project codes, creating confusion in compliance records reviewed by the Delhi Development Authority.

The National Informatics Centre, which manages backend infrastructure for several Delhi government portals including the Delhi e-District platform at Lodi Road, began deploying perceptual hashing tools — software that generates a fingerprint for each image and flags matches above a set similarity threshold — across relevant datasets on June 30. The technology is not new; it has been in use for content moderation on social platforms for years. Applying it to legacy government archives at this scale, however, requires manual review of every flagged pair before deletion, because an automated purge risks removing legitimate records that happen to share visual similarity.

What Residents and Applicants Should Watch For

Anyone who applied for a caste certificate, domicile document, or driving licence renewal through Delhi's online portals between January 2021 and March 2024 may find their application record temporarily inaccessible if the photograph attached to it has been flagged for review. The Department of Information Technology advised citizens to carry physical copies of their documents to service centres at the Pragati Maidan integrated transit complex and at Citizen Service Bureaus in Dwarka Sector 10 and Laxmi Nagar during the review window, which runs until July 31.

Civic groups in Karol Bagh and Patparganj have already raised concerns that lower-income residents who rely on government-issued photo identity for everything from ration entitlements to hospital admission could face unnecessary hurdles if their records get caught in the deduplication sweep without adequate notice. The IT department has not yet announced a specific grievance redressal window, though administrators have indicated that a dedicated helpline number will be published before July 10. The cleanup may be technically necessary — but for the roughly 14 million voters whose records sit in the affected databases, the proof will be in how smoothly the process lands.