Q-PIPE A Practical Quantum Phase Encoding Method

Brian García Sarmina, Emmanuel Martínez-Guerrero, Janeth De Anda Gil, Sun Guo-Hua, Dong Shi-Hai · 2026

Read Q-PIPE A Practical Quantum Phase Encoding Method by Brian García Sarmina, Emmanuel Martínez-Guerrero, Janeth De Anda Gil, Sun Guo-Hua, Dong Shi-Hai (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-09869.txt.

Introduction

García Sarmina and four co-authors present Q-PIPE, a method loading standard data into quantum states, writing values into oracle phases, turned by phase guessing. Sums and gaps come from stepwise oracle calls, with no quantum math circuits. Applied to image edge finding by aimed finite gaps, the oracle needs gated-NOT gates scaling with pixels and guessing qubits, matching tuned novel boosted quantum form setup. Evenly gated build cuts depth 53 to 59 times and preserves edge structure under device noise, with match 0.99 against the standard gradient.

Catalog record

CREATORBrian García Sarmina, Emmanuel Martínez-Guerrero, Janeth De Anda Gil, Sun Guo-Hua, Dong Shi-Hai
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE53198
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.09869

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Q-PIPE A Practical Quantum Phase Encoding Method Brian García Sarmina, Emmanuel Martínez-Guerrero, Janeth De Anda Gil, Sun Guo-Hua, Dong Shi-Hai Abstract A major hurdle in Quantum Image Processing (QIMP) is the efficient transfer of classical, high-dimensional image data into quantum states. Current methods present a trade-off: amplitude encoding (FRQI) is computationally expensive in terms of quantum gates and limited arithmetic operations, while basis encoding (NEQR) has a significant initialization overhead that scales with both image resolution and intensity bit-depth. Moreover, current frequency-domain approaches require complex transformations for basic pixel-wise arithmetic or extensive post-processing to retrieve the original (or processed) pixel information. To address the lack of practical phase encoding methods, we introduce Q-PIPE (Quantum-Gray Phase Injection for Pixel Encoding). By exploiting the quantum phase kickback mechanism and optimizing spatial traversal via a Gray-code sequence,Q-PIPE efficiently maps continuous intensity values into the computational basis with an optimized elementary gate count of O ⁡ ( q ​ N ) O(qN) , a O ⁡ ( log ⁡ N ) O(\log N) improvement over standard basis encoding. This architecture utilizes the phase domain as the operational space, allowing for the native computation of finite differences without deep arithmetic circuits. We systematically address classical readout vulnerabilities, such as phase aliasing and spectral leakage, by mapping inputs to a [ − π , π ] [-\pi,\pi] domain and introducing a probability threshold equation that scales inversely with the dimension of the spatial register. A proof-of-concept implementation performing Quantum Edge Detection (QED) via directional derivatives demonstrates strong accuracy, achieving exact reconstruction for quantized inputs and low Mean Absolute Error (MAE) for continuous data across multiple benchmark datasets. Ultimately, Q-PIPE establishes a highly parallelizable, NISQ-compatible subroutine that not only advances quantum computer vision, but also offers strong potential to mitigate input/output (I/O) data-loading overhead in broader Quantum Machine Learning (QML) applications. keywords Quantum Phase Encoding, Quantum Phase Estimation, Quantum Fourier Transform, Quantum Image Processing. pacs 03.67.Ac; 03.67.Bg; 03.65.Ta † † equal-contributors: All authors contributed equality to this work. † † equal-contributors: All authors contributed equality to this work. † † equal-contributors: All authors contributed equality to this work. † † equal-contributors: All authors contributed equally to this work. 1 Introduction The rapid advancement of quantum computing has catalyzed the emergence of Quantum Image Processing (QIMP) as a promising field, driven by the potential to achieve exponential speedups over classical counterparts in storage and processing tasks 70 ; 25 ; 17 ; 50 . As quantum hardware continues to advance, the central challenge has shifted from theoretical possibility to efficient implementation of high-dimensional data handling; particularly, the fundamental bottleneck of encoding classical image data into quantum states 34 ; 30 ; 65 ; 25 ; 73 . Current Quantum Image Representation (QIR) models generally fall into a dichotomy defined by a fundamental trade-off between spatial efficiency and operational tractability 49 ; 48 ; 42 . On one hand, methods such as the Flexible Representation of Quantum Images (FRQI) 31 ; 28 ; 40 ; 26 utilize probability amplitudes to encode (amplitude encoding) pixel intensity. While FRQI achieves a logarithmic qubit complexity of O(\log N) 32 ; 37 , this spatial compression incurs a prohibitive computational overhead during state preparation, demanding an O(N^{2}) gate count and circuit depth. Its dependence on entangled continuous amplitudes makes arithmetic operations and pixel processing computationally expensive and complex to implement. On the other hand, the Novel Enhanced Quantum Representation (NEQR) 74 ; 40 ; 15 ; 51 encodes pixel intensity values directly in the computational basis states. This basis encoding structure significantly simplifies subsequent processing steps, such as color transformations and thresholding, since digital information is high accessible 51 ; 60 . However, this accessibility comes at a high initialization cost. At the elementary gate level, the necessary multi-controlled operations reveals that the loading phase for standard NEQR demands a gate count and circuit depth scaling of O(qN\log N) , where q is the intensity bit-depth. To bypass these limitations, recent literature has increasingly explored encoding images in the frequency or phase domains 34 ; 39 ; 58 ; 16 ; 23 . Strategies such as Fourier Transform representations 20 ; 65 ; 12 ; 55 , multi-layered hyperspectral representations 3 , and Cosine Series Approximations (QCoSamp) 67 ; 65 map image information into spectral coefficients. More recently, a framework for mapping images directly onto the unit circle (using the phase domain) have been proposed 21 . Although these approaches successfully compress image data by treating it as a signal, they often require complex transformations to perform basic pixel-wise operations or complex (and highly computationally demanding) post-processing methods, making the data hard to manipulate. Another line of research emphasizes the potential of phase-oriented representations through the development of phase-based encoding schemes, such as Local Phase Image Quantum Encoding. In this framework, pixel information is explicitly embedded into the phase of quantum states, enabling a representation that is inherently compatible with phase-sensitive quantum operations. This characteristic makes such approaches particularly well-suited for photonic quantum systems, where phase manipulation constitutes a natural and experimentally accessible resource 68 .However, these methods typically suffer from significant challenges in information retrieval, as phase-encoded data is not directly accessible through standard measurements and often requires complex interference-based procedures. This introduces additional circuit overhead and increases susceptibility to noise, potentially leading to information loss during readout. Other recent advances in quantum image representation have focused on improving scalability and operational efficiency. Contemporary approaches investigate alternative encoding strategies designed to reduce circuit depth, optimize resource utilization, and enhance robustness against noise, thereby improving overall fidelity in practical implementations 22 ; 35 . Despite these improvements, such approaches often involve non-trivial trade-offs, where gains in scalability or noise resilience come at the cost of increased encoding complexity, limited flexibility for pixel-wise operations, or additional constraints on circuit design, which may hinder their applicability in more general image processing tasks. In this work, we address this efficiency-complexity trade-off by proposing an encoding method that retains the operational advantages of basis encoding while substantially reducing the preparation overhead. The theoretical foundation of phase encoding is well established 44 ; 43 . Our approach strategically combines three main ideas: the phase kickback within a framework similar to Quantum Phase Estimation, state marking and unmarking akin to Grover’s algorithm combined with a Gray-code representation, and post-processing considerations to improve the algorithm’s performance (and overcome certain known problems), which creates a practical quantum phase encoding algorithm: the Quantum-Gray Phase Injection for Pixel Encoding (Q-PIPE). By framing the image loading as a parameter estimation problem, we exploit the quantum phase kickback mechanism 44 ; 72 , to systematically map continuous intensity values onto the relative phase, and subsequently project them into the target computational registers. This protocol offers a precise mathematical pathway to encode data, connecting the efficiency of phase-domain representations with the operability of basis-encoded methods. The remainder of this paper is organized as follows: Section 2 introduces the preliminary concepts of global and relative phases, alongside the foundational QPE algorithm. Section 3 formally details the proposed Q-PIPE encoding architecture and its step-by-step circuit evolution. Section 4 extends the framework to multi-image encoding, demonstrating how arithmetic operations like phase accumulation occur naturally during the loading stage. Section

Read the complete text