A quantum circuit library for image processing

Today’s interest in applications like pattern recognition and computer vision makes image processing algorithms of great importance. However, the rapidly increasing volume of visual information weighs on the computational capabilities currently available in classical computers. Quantum Image Processing (QImP) focuses on providing a counterpart of conventional image processing strategies in the Quantum Computing domain, exploiting its intrinsic parallel nature. Over the years, many QImP algorithms have been proposed to encode and process images using quantum formalism. Despite this, in the state-of-the-art, not enough room is given for direct and practical comparisons between the available techniques. Therefore, difficulties arise when trying to understand whether they represent effective opportunities with respect to classical counterparts, especially when considering the limitations and non-idealities of nowadays quantum hardware. The goal of this thesis is to define a Python software library of QImP algorithms compatible with Qiskit, an open-source software-development kit for quantum computing, to provide users with the ability to flexibly compare the different techniques on reference input images and analyze their suitability through particular figures of merit. First, a preliminary study of the current literature on QImP has been carried out in order to identify the most promising algorithms. Then, they have been implemented as parametrical modules, which progressively formed the library. Jupyter Notebooks were considered to provide a practical user guide for a conscious application of the algorithms provided. The selection of the supported techniques spans from encoding methods, basic processing tools, compression, and edge detection algorithms and takes into account the limited computational resources of quantum hardware and the possibility of practical applications. Tests have been conducted on all the implemented circuits, both through simulations on classical computers and tests on real quantum hardware. The strengths and weaknesses in the application of the different algorithms have been put into evidence considering several use cases. This thesis lays the foundations for exploring the QImP scenario, while the implemented library gives the possibility to include and characterize new algorithms and compare them with the others, thanks to its flexible and modular nature.