Memristive Crossbar Array And Its Application On Image Enhancement And Storage With Computing

As more research is devoted to the human visual system,using the visual processing mechanism of human eyes to enhance the visual cognition of the machine has become a hot spot in the direction of computer vision.In the early experiments,we found a nonclassical receptive field on the retina of the human eye.Its structure is a concentric circle,which can detect the brightness and edge information of the image.Then the three Gauss model has achieved a certain effect on the image enhancement,but because of its excessive parameters,it is too tedious for a large number of pictures.The memristor is a kind of resistance element with bionic characteristics.Its nonvolatile,low power,nanoscale and other excellent properties make it play an extraordinary role in the fields of cross array storage,logic operation and image processing.Making use of the characteristics of the memristor to dynamically change some of the parameters in the three Gauss model can achieve very good results in image enhancement.In addition,in recent years,microelectronic chips have developed rapidly and their integration and performance have been improved rapidly.However,the traditional storage technology has gradually come to a physical bottleneck,which leads to the more and more serious problems of the computer's "storage wall"(the gap between processors and memory).The birth of memristor brings infinite vision for the researchers who manufacture computers with larger storage capacity and faster processing speed.The memristor is nonvolatile,and can be used as a neural synapse with its natural memory characteristics,even when it is broken down.It is expected to break the gap between the processing and storage of the Von Neumann architecture by simulating the rules of STDP and SRDP learning in the biological system.This paper analyzes the memristive cross array storage and its application in image processing and the STDP and SRDP rule based on memristor synapses.Then,combining the traditional three Gauss model,a new adaptive three Gauss model is proposed,which is combined with the memristive cross array structure,and is better in grayscale image and color image enhancement.Finally,a novel type of memristive cross array is designed,which achieves the integration of image storage and operation by memristive synaptic plasticity.First of all,this paper introduces a lot of research on the emergence of memristor in the field of image processing.Two typical memristor models are analyzed,and the model derivation and numerical simulation are carried out respectively.The storage principle of the memristor and the application analysis of the cross array structure in the image storage and output are deeply studied,and the effectiveness is verified by experiments.Then,a new adaptive three Gauss model is proposed.On the basis of the classic three Gauss model,combined with some properties of the memristive cross array structure,the parameters in the traditional three Gauss model can be dynamically changed,thus the local image enhancement is optimized,and the overall input image can have better enhancement effect.It is necessary to calculate the typical local features of the input image,which can be used to determine the voltage characteristics at the two ends of the memristor which is at the pixel point,so that the memristor is changed accordingly.Next,the value of the parameters in the corresponding model is calculated according to the recalling value obtained,thus the enhancement template of the local sub image in the input image can be determined to complete the adaptive image enhancement.Finally,a silver sulfide memristor is used as a biological synapse to simulate biological systems and STDP and SRDP learning rules.On this basis,a new type of cross array structure of memristor is designed so that the structure can integrate storage and operation together,and use the memristive synaptic plasticity to adjust the resistance of the memristor to complete a series of basic operations of the image.This new type of memristive cross array and image processing are combined to realize the basic operation of image processing.It is hopeful to break the gap between processing and storage in the Von Neumann architecture.