A novel approach is proposed for multiband image processing via quantum models in real situations. Quantum circuits are automatically generated ad-hoc for each use case via multiobjective genetic algorithms. Using this universal method, image processing tasks such as segmentation can be carried out by considering the properties that constitute each pixel. The generated circuits present a low level of correlation between qubits, and thus can be considered quantum-inspired machine learning models. The effectiveness of this methodology has been validated by applying it to different segmentation use cases. Comparisons are made between optimized classical kernel methods and the generated quantum-inspired models to understand their behaviors. The results show that quantum models for multiband image processing achieve accuracies similar to those of classical methods. • Novel method for automatically generating quantum models for multiband images. • Automatic generation of quantum models via genetic algorithms. • Quantum-inspired machine learning models for pixel segmentation. • Real-world applications of quantum models. • Quantum circuits handling images.
A novel approach is proposed for multiband image processing via quantum models in real situations. Quantum circuits are automatically generated ad-hoc for each use case via multiobjective genetic algorithms. Using this universal method, image processing tasks such as segmentation can be carried out by considering the properties that constitute each pixel. The generated circuits present a low level of correlation between qubits, and thus can be considered quantum-inspired machine learning models. The effectiveness of this methodology has been validated by applying it to different segmentation use cases. Comparisons are made between optimized classical kernel methods and the generated quantum-inspired models to understand their behaviors. The results show that quantum models for multiband image processing achieve accuracies similar to those of classical methods. • Novel method for automatically generating quantum models for multiband images. • Automatic generation of quantum models via genetic algorithms. • Quantum-inspired machine learning models for pixel segmentation. • Real-world applications of quantum models. • Quantum circuits handling images.
Quantum machine learning (QML) merges quantum computing and machine learning to enhance data processing. QML models typically use quantum circuits, but designing these circuits is challenging due to issues with handling large data volumes and circuit size. Larger circuits can be inefficient, while smaller ones may lack expressivity. This research introduces a global optimization framework for generating quantum circuits with multi-objective genetic algorithms, optimizing structure, operators, and parameters while balancing accuracy and model size. The gradient-free optimization method avoids issues like barren plateaus. The proposed methodology is applied in various domains, including classification, image segmentation, and time-series prediction, achieving promising results. The discovered models have high representational capacity with minimal qubit use, improving scalability and efficiency for large models and inspiring quantum-inspired machine learning models for both academia and industry. RESUMEN El aprendizaje automático cuántico combina la computación cuántica y el aprendizaje automático para mejorar el procesamiento de datos. Los modelos suelen utilizar circuitos cuánticos, pero su diseño es un reto debido a los problemas que plantea el manejo de grandes volúmenes de datos y el tamaño de los circuitos. Los circuitos más grandes pueden ser ineficientes, mientras que los más pequeños pueden carecer de expresividad. Esta investigación introduce un marco de optimización global para generar circuitos cuánticos con algoritmos genéticos multiobjetivo, optimizando la estructura, los operadores y los parámetros al tiempo que se equilibra la precisión y el tamaño del modelo. El método de optimización sin gradiente evita problemas como barren plateaus. La metodología propuesta se aplica en diversos ámbitos, como la clasificación de datos tabulares e imágenes, la segmentación de imágenes y la predicción de series temporales, obteniéndose resultados prometedores. Los modelos descubiertos tienen una alta capacidad de representación con un uso mínimo de qubits, lo que mejora la escalabilidad y la eficiencia de los modelos de gran tamaño e inspira modelos de aprendizaje automático de inspiración cuántica tanto para el mundo académico como para la industria.
Abstract Image segmentation remains one of the most extensively studied problems in computer vision, with applications ranging from medical image analysis and autonomous driving to remote sensing and industrial inspection. While deep learning approaches currently dominate the field, interest in quantum computing has motivated the development of quantum and quantum-inspired segmentation techniques that explore alternative strategies for image representation, optimization, and learning. This survey provides a comprehensive review of quantum and quantum-inspired image segmentation methods within the broader context of modern segmentation research. To establish this context, we first examine the evolution of image segmentation methodologies, covering classical image processing techniques, machine learning approaches, and contemporary deep learning architectures.We then organize the existing quantum image segmentation literature into four categories: quantum-inspired, tensor network-based, hybrid quantum-classical, and fully quantum methods. Across these categories, we analyze computational principles, image encoding and representation strategies, application domains, and datasets. We further compare quantum and quantum-inspired approaches with established segmentation paradigms in terms of supervision, scalability, and hardware requirements. Our analysis indicates that quantum-inspired methods currently represent the most developed direction, while hybrid approaches are emerging and fully quantum methods remain constrained by quantum hardware, data encoding costs, and scalability challenges. By connecting quantum image segmentation with the broader evolution of segmentation research, this survey provides a unified perspective on the field. It identifies current trends, open challenges, and promising future directions for researchers exploring the emerging role of quantum computing in image segmentation.
Abstract We propose a new technique for the automatic generation of optimal ad-hoc ansätze for classification by using quantum support vector machine. This efficient method is based on non-sorted genetic algorithm II multiobjective genetic algorithms which allow both maximize the accuracy and minimize the ansatz size. It is demonstrated the validity of the technique by a practical example with a non-linear dataset, interpreting the resulting circuit and its outputs. We also show other application fields of the technique that reinforce the validity of the method, and a comparison with classical classifiers in order to understand the advantages of using quantum machine learning.
A new hybrid system is proposed for automatically generating and training quantum-inspired classifiers on grayscale images by using multiobjective genetic algorithms. It is defined a dynamic fitness function to obtain the smallest circuit complexity and highest accuracy on unseen data, ensuring that the proposed technique is generalizable and robust. At the same time, it is minimized the complexity of the generated circuits in terms of the number of entangling operators by penalizing their appearance and number of gates. The size of the images is reduced by using two dimensionality reduction approaches: principal component analysis (PCA), which is encoded within the individual and genetically optimized by the system, and a small convolutional autoencoder (CAE). These two methods are compared with one another and with a classical nonlinear approach to understand their behaviors and to ensure that the classification ability is due to the quantum circuit and not the preprocessing technique used for dimensionality reduction.
Machine learning is the process of discovering patterns in data and using those patterns to forecast or predict future occurrences of those patterns. The machine learning process consists of three steps: input data management, model construction, and model evaluation. Combining machine learning and quantum computing allows for the design and operation of quantum circuits with behavior resembling that of classical machine learning. Variable quantum circuits are frequently used in quantum machine learning. Parameterized variables, a crucial part of variable quantum circuits, must be adjusted in order to achieve good performance of these circuits. It is possible to modify parameterized variables in a variety of ways, such as by using gradient methods or linear programming to determine the circuit’s optimal point. This study proposes a method for automatically designing a variational quantum circuit to classify the different datasets, including the well-known 1-dimensional and 2-dimensional datasets, iris and MNIST, using a genetic algorithm to find the best values for the parameterized variables. The Qiskit quantum simulation tools are employed in these experiments. The results show that genetic algorithms can find solutions with greater accuracy than linear-based optimizers by 94 percent and 37.5 percent, respectively, for the iris dataset and the MNIST dataset.
Quantum computing is expected to transform a range of computational tasks beyond the reach of classical algorithms. In this work, we examine the application of variational quantum algorithms (VQAs) for unsupervised image segmentation to partition images into separate semantic regions. Specifically, we formulate the task as a graph cut optimization problem and employ two established qubit-efficient VQAs, which we refer to as Parametric Gate Encoding (PGE) and Ancilla Basis Encoding (ABE), to find the optimal segmentation mask. In addition, we propose Adaptive Cost Encoding (ACE), a new approach that leverages the same circuit architecture as ABE but adopts a problem-dependent cost function. We benchmark PGE, ABE and ACE on synthetically generated images, focusing on quality and trainability. ACE shows consistently faster convergence in training the parameterized quantum circuits in comparison to PGE and ABE. Furthermore, we provide a theoretical analysis of the scalability of these approaches against the Quantum Approximate Optimization Algorithm (QAOA), showing a significant cutback in the quantum resources, especially in the number of qubits that logarithmically depends on the number of pixels. The results validate the strengths of ACE, while concurrently highlighting its inherent limitations and challenges. This paves way for further research in quantum-enhanced computer vision.