In the context of our focus on »Quantum Image Processing« we investigate to what extent quantum computers (QC) can solve classical image processing problems. In addition to the theoretical results, we focus on the practical implementation on current hardware and the identification of new possibilities.
We start from conventional images, thus the classical image information has to be converted into quantum states. In order to explore the current possibilities, we performed experiments on both, quantum computer simulators and real backends: Classical images were converted to quantum states and back into classical images. In these experiments, we succeeded in practically demonstrating the theoretically promised advantage of exponentially fewer qubits (quantum bits) compared to classical bits.
However, full exploitation of this advantage is hampered by the flaws of current quantum computers, which strongly limit the executability of algorithms. Currently, images of maximum size 16 by 16 pixels can be processed, but only two by two pixel gray value images are still interpretable after a cycle of coding and decoding. Larger images are noisy beyond recognition. For details please check the paper »Improved FRQI on superconducting processors and its restrictions in the NISQ era« by Geng et al. (see below).
Our improvements of the encoding method reduce errors and memory consumption and increase the size of processable images to 32 by 32 pixels. Currently, we focus on developing hybrid image processing algorithms making use of the quantum computing advantages in spite of this size restriction.