Memristor based Crossbar Array using CNN for Image Processing Applications

Image processing has become one of the popular neuromorphic computing applications in recent years. Most of the application areas require an efficient and real time image processing system. Memristor is a useful memory device which plays an important role to design an image processing system and realization of hardware design. Due to the non-volatile nature of memory and less power consumption, Memristor has a huge application in the field of image processing. Being a memory device, Memristor can preserve data and parameters during processing time and reducing the need for data transfer as well as storage. In this research, Memristor in the form of CNN architecture will store the image extracted data where pixel image is given as an input and voltage pulses are generated according to the pixel value. The voltage pulses are then applied as input to the Memristor crossbars and compared with a certain threshold voltage above which an output neuron responds. The above mention system is trained for image recognition applications in real time scenarios. The real time image processing based Memristor system can have huge application areas such as medical imaging, security, computer vision and robotics.