Extracting Quantum-Ready Cores in Discrete Tomography
Year: 2026
Type: conference-paper
Abstract: We introduce a structure-aware hybrid classical–quantum framework for discrete binary tomography from very few projections that enables practical quantum speedups through an explicit reduction of the problem’s combinatorial complexity. Rather than directly applying quantum algorithms to the full reconstruction task, we reshape the reconstruction problem by combining continuous relaxation and constraint propagation to deterministically fix the majority of pixel values, thereby isolating a small uncertain core. This core captures the residual combinatorial structure and has size k = O(n) for n × n images, reducing the effective search space from \(2^{n^2}\) to 2k, where k = O(n) and typically k ≪ n2 in practice. This reduced formulation is inherently compatible with quantum search. In particular, the refinement phase can be mapped to a search over {0, 1}k, enabling quadratic speedups via Grover’s algorithm, as well as more refined quantum backtracking strategies such as Montanaro’s algorithm. Experimental results on four projections from binary images up to 512 × 512 pixels show that the classical component alone achieves near-linear scaling and high reconstruction accuracy (above \(99.5\%\)), significantly outperforming DART and remaining competitive with ILP approaches. Importantly, the combinatorial refinement is both sparse and selective, activating primarily on difficult instances and exploring only a small fraction of the reduced search space. These findings demonstrate that the proposed decomposition acts as an effective structure-extraction layer, bridging classical reconstruction and quantum search, and providing a concrete pathway toward scalable and practically relevant quantum-enhanced discrete tomography.
Authors: Simone Faro, Arianna Pavone, Cesare Valenti
Institutions: University of Catania, University of Palermo
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7168021217