Quantum and Quantum-Inspired Image Segmentation in Computer Vision: Bibliometric and Scientometric Analysis of Research Trends and Hybrid Methods

Quantum and quantum-inspired image segmentation lies at the intersection of quantum computing, computer vision, machine learning and medical image analysis. We map the field from 422 Scopus-indexed publications (2018–2025) retrieved with a documented Advanced Search query and screened in three explicit levels. Formal eligibility retained 408 records. A published topical decision rule was then applied in two operationalizations that differ only in the metadata scope they read (observed agreement 82.8%, Cohen’s κ = 0.65); all 197 records in the two agreement-relevant strata were adjudicated at record level, and the 211 records both operationalizations excluded were validated on a random sample of 60. The final topical corpus contains 143 records, and at the sample rate roughly two to three further eligible records remain in the unexamined remainder. Screening changes the headline picture: the retrieved corpus grows at 20.45% a year (28 records in 2018 to 103 in 2025) but the topical corpus at only 10.41% (13 to 26), so about half of the apparent expansion is peripheral retrieval. China, India and the United States lead at every level. Of the 138 method studies, 9 (6.5%) report execution on physical quantum hardware, 43 (31.2%) are gate-based or variational studies run on simulators or with an unstated environment, 12 (8.7%) use annealing, Ising or QUBO formulations, and 74 (53.6%) are quantum-inspired classical algorithms that require no quantum device; 27 (19.6%) combine a classical vision pipeline with a quantum component. Quantum-inspired classical work falls from 78.0% to 29.2% of the period’s method studies while hybrid work rises from 3 to 19 records. All findings describe publication patterns, not segmentation performance or quantum advantage. A versioned data and code package reproduces every reported number.

Hali tarjima qilinmagan