Improved Image Feature Extraction Based on Quantum Convolution

Quantum computing has proliferated by offering the potential for significant exponential speed improvements. Quantum computing potential can be exploited in various fields, such as complex collision image processing in traffic surveillance. Quantum-based computing performs data processing through the principles of quantum mechanisms such as superposition and entanglement. Quantum computing has various applications, including deep learning. The convolution process in deep learning methods can be modified using quantum circuit mechanisms and promises to increase the efficiency of feature extraction compared to classical methods. This paper explains the concept of quantum convolution by exploring two main mechanisms, i.e., entire quantum circuits for convolution and quantum circuits as quantum kernels. We also propose a new quantum circuit that can represent the convolution process. Theoretically, quantum convolution is faster than classical computation. Our results advance knowledge and provide new insights into developing better convolution for feature extraction.