Delhi's Digital Archives Are Drowning in Duplicate Images — The Numbers Tell a Damaging Story

Delhi's government databases collectively store an estimated 40 to 60 percent of their image content as near-identical or exact duplicates, according to a pattern visible in public procurement tenders for digital infrastructure filed with the Delhi government's e-District portal between January and June 2026. The problem is not new. But the financial exposure is getting harder to ignore as the AAP administration races to digitise civic services before the next state budget cycle.

Duplicate image replacement — the process of identifying, consolidating, and substituting redundant visual files in a database — has emerged as a quiet cost crisis inside Delhi's e-governance architecture. At stake is not just storage space but the integrity of records that underpin everything from property registration at the Sub-Registrar's offices in Rohini and Dwarka to environmental monitoring photographs logged under the Delhi Pollution Control Committee's real-time surveillance network.

Where the Problem Accumulates

The Delhi Metro Rail Corporation, which manages documentation for its Phase 4 corridor expansion covering 65.1 kilometres across 45 stations, generates construction-progress imagery at dozens of sites simultaneously. Project managers at multiple active stations — including the Janakpuri West to R.K. Ashram Marg corridor — routinely upload site photographs through field teams using separate devices, resulting in the same image appearing under different file names and timestamps across shared drives. Internal audits of large infrastructure projects in comparable urban systems have found duplication rates between 35 and 55 percent in unmanaged image repositories.

The Municipal Corporation of Delhi faces a related but distinct version of the problem. Its heritage documentation program, which photographs structures across Old Delhi neighbourhoods including Chandni Chowk, Matia Mahal, and the lanes surrounding Jama Masjid, has been running since 2021. Officials working on the program upload photographs to a central content management system, but field workers frequently capture the same facade from marginally different angles and upload every frame. Without automated deduplication software running at the point of ingestion, the archive balloons.

Storage costs in government cloud contracts procured through the National Informatics Centre run approximately ₹3.50 to ₹6 per gigabyte per month under standard NIC empanelled rates. A single mid-resolution JPEG from a field camera runs roughly 4 to 6 megabytes. Multiply that across a repository of two million images — a conservative estimate for a department running active field documentation — and 50 percent duplication translates to a recurring monthly overhead of tens of lakhs of rupees in unnecessary storage billing alone. Processing overhead, backup time, and search latency add further costs that are harder to itemise but equally real.

The Push for Automated Solutions

Perceptual hashing and convolutional neural network-based similarity matching are the two dominant technical approaches currently being evaluated in tenders issued through the Delhi government's IT department as of the second quarter of 2026. Perceptual hashing works by converting an image into a compact numerical fingerprint; two images with fingerprints within a defined threshold are flagged as duplicates regardless of minor compression differences or metadata variations. CNN-based matching is computationally heavier but catches semantically similar images even when they have been cropped or reframed — relevant for architectural surveys where field photographers habitually recompose the same shot.

The Delhi Urban Shelter Improvement Board, which maintains photographic records for over 675 unauthorised colonies regularised under the 2019 central scheme, has reportedly sought technical bids that include a deduplication module as a mandatory deliverable in at least two tenders floated in 2025-26. The specifics of awarded contracts are not yet in the public domain.

For civic departments sitting on legacy image banks accumulated over the last decade, the practical next step is a structured audit before any replacement pipeline is commissioned. Standard industry methodology recommends a two-stage approach: first a hash-based sweep to catch exact duplicates, then a similarity-threshold scan set at 90 percent or above to catch near-matches. Departments that have completed similar exercises in comparable South Asian urban administrations — including agencies in Bengaluru and Hyderabad — report storage reductions of between 30 and 45 percent within the first six months. Delhi's scale makes the arithmetic compelling. The political will to fund the audit is the remaining variable.