Quantum-enhanced Computer Vision: Going Beyond Classical Algorithms

TL;DR

Quantum-enhanced Computer Vision (QeCV) investigates how quantum computing can accelerate and augment vision tasks by exploiting the two primary paradigms gate-based quantum computing and adiabatic quantum computing. The paper provides a comprehensive survey of problem mappings, high-level methodologies, and the state of hardware, frameworks and learning materials, emphasizing experimental evaluation on real quantum devices where possible. It highlights that most practical CV quantum algorithms are formulated as QUBO/Ising optimizations or PQCs and often operate in hybrid classical–quantum pipelines, with iterative, hardware-aware workflows. Despite substantial progress, the field faces challenges in hardware noise, scalability, data encoding, and reliable benchmarking, but ongoing advances such as quantum error correction milestones and increasingly capable simulators point to a promising long-term impact on CV practice and research.

Abstract

Quantum-enhanced Computer Vision (QeCV) is a new research field at the intersection of computer vision, optimisation theory, machine learning and quantum computing. It has high potential to transform how visual signals are processed and interpreted with the help of quantum computing that leverages quantum-mechanical effects in computations inaccessible to classical (i.e. non-quantum) computers. In scenarios where existing non-quantum methods cannot find a solution in a reasonable time or compute only approximate solutions, quantum computers can provide, among others, advantages in terms of better time scalability for multiple problem classes. Parametrised quantum circuits can also become, in the long term, a considerable alternative to classical neural networks in computer vision. However, specialised and fundamentally new algorithms must be developed to enable compatibility with quantum hardware and unveil the potential of quantum computational paradigms in computer vision. This survey contributes to the existing literature on QeCV with a holistic review of this research field. It is designed as a quantum computing reference for the computer vision community, targeting computer vision students, scientists and readers with related backgrounds who want to familiarise themselves with QeCV. We provide a comprehensive introduction to QeCV, its specifics, and methodologies for formulations compatible with quantum hardware and QeCV methods, leveraging two main quantum computational paradigms, i.e. gate-based quantum computing and quantum annealing. We elaborate on the operational principles of quantum computers and the available tools to access, program and simulate them in the context of QeCV. Finally, we review existing quantum computing tools and learning materials and discuss aspects related to publishing and reviewing QeCV papers, open challenges and potential social implications.

Figures (25)

• Figure 1: Quantum- enhanced computer vision. . (A): First, a target problem must be formulated in a form consumable by modern quantum machines, e.g. as a QUBO problem for AQC devices or as a gate sequence for gate-based QC. This operation is performed on a host (classical CPU). (B): In AQC, the resulting QUBO defines a logical problem—binary variables that become qubits during optimisation on an idealised quantum annealer with full qubit connectivity. Alternatively, gate-based QC uses a gate sequence to drive the system into a solution-encoding state. (C): To run on a quantum computer with limited connectivity, a logical problem must be minor-embedded or transpiled. During this mapping step, each logical qubit is assigned to one or more physical qubits to match hardware constraints. (D): An AQC device performs annealing for computation, while a gate-based QC device alternatively executes the algorithm describing gates. Adiabatic computers leverage quantum mechanical effects of superposition and tunnelling to find optima of QUBOs. Gate-based computers can additionally harness entanglement and interference to speed up computations, surpassing the capabilities of classical ones. (E): Measured qubit values are unembedded from the hardware and aggregated in the AQC paradigm, or directly read out in gate-based QC. The measurement is repeated several times, and a solution distribution is returned to the host. The bit-strings are processed and interpreted in terms of the original problem. Image sources, if applicable (from left to right and top to bottom in each step): (A): QuantumSync2021, huang2021multibodysync, xiang2015learning, ransac, LiGhosh2020, krizhevsky2009learning, yan2023toward, cong2019quantum, (B): QuantumSync2021, meyer2022survey, yang2024robust, (C): yurtsever2022q, ionq_aria, (D): dwave_wiki, sycamore_wiki.
• Figure 2: Visualising an arbitrary state of a qubit $\ket{\psi}$ on the Bloch sphere along with its several widely encountered states. Although in the original space $\mathbb{C}^2$ the states $\ket{0}$, $\ket{1}$ are orthogonal, they are visualised as opposite poles on the Bloch sphere.
• Figure 3: Common processing stages in a typical quantum circuit. Here, the quantum circuit operates on $n$ qubits, which at the beginning are often initialised to a basic state, e.g. $\ket{\psi_1 \psi_2 \dots \psi_n} = \ket{10\dots 0}$. Then, the basic state is prepared to yield an initial (usually superimposed) state, before a sequence of $M$ quantum gates is invoked. The specific gates employed define the algorithm implemented. The final quantum state is then measured to produce classical outcomes that correspond to the final results, which are often probabilistic. It is also common for a circuit to be repetitively invoked or iterated.
• Figure 4: Conceptual illustration of AQC. (Right): Depiction of the spectral gap as the smallest difference between the lowest and the first excited energy states of the system over the time $t$. (Left): The AQC process of minimising a function $\braket{x|H(t)|x}$ that is interpolated from a trivial initialisation objective to a QUBO problem of interest. The minimum eigenvalue on the left corresponds to a superimposed (eigen)state that determines the measurement probabilities of the basis states on the right. For a strictly positive spectral gap, a sufficiently slow transition process maximally boosts the likelihood of measuring the ground state of the QUBO.
• Figure 5: Diagram of QAOA. The Hamiltonian $H_C$ in the figure corresponds to $H_P$ in this survey. Image source: PhysRevX.10.021067.

Theorems & Definitions (14)

• theorem 1
• proof
• theorem 2
• proof
• theorem 3
• proof
• lemma 1
• proof
• proof
• theorem 4