Font Size: a A A

Research On Image Processing Based On Memristors

Posted on:2015-03-19Degree:DoctorType:Dissertation
Country:ChinaCandidate:J ZhouFull Text:PDF
GTID:1108330479479622Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
With the development of computer technology, digital image processing has been used widely in many fields, such as engineering, computer science, information science,statistics, physics, chemistry, biology, medicine, military and even social science, and it has become a practical technology closely related to the nation’s development and people’s daily life. However, as digital image processing is a type of memory-access intensive application, with large amount of image data and complex algorithm, how to enhance the efficiency becomes a plague of the image processing field. Especially in Information Age, with the popularization of Internet and the coming of Big-Data Era, great number of image data need realtime processing efficiently, which urges us to research efficient realtime image processing techniques.Memristor is a new kind of enable device found recently. It has several excellent properties, such as nonvolatility and high- density, which makes it a promising candidate for future memory; thus nanomemory based on memristor becomes a hotspot in recent years. Researchers have also found that, as a memory architecture, memristor crossbar array memory structure could be used to compete logic operations. This illuminates that memristors has the ability of both storage and computation, and can complete in situ logic operation(logic operation is executed where the operations are stored), which provides a hopeful technique to solve the problems of image processing thus worth being researched.With the ability of memristor of combining storage and computation, this thesis aims at researching image processing techniques based on memristors after analysis of the status quo and problems of image processing. Firstly, we analyze the image processing algorithms and propose the architecture combining image processing and storage based on memristor; secondly, we design the basic image operations in this architecture as well as several common basic logic elements; Thirdly, we validate our design with experiment and performance analysis.Specifically, the main work and contributions of this paper are as follows:1. Proposing the architecture combining image processing and storage based on mem-ristors(Chapter 2)After the analysis of image processing algorithms, we propose the memristor-based architecture combining image processing and storage. First of all, we investigate thealgorithms in image processing, and choose the ones suited to memristor. Then, webasic image operations and analyze their performance for thereafter archi-tecture design and performance comparison. With the analysis above, we propose the architecture combining image processing and storage based on memristors, in-cluding basic idea, the storage of image data and the read/write of image data. In this combining architecture, image operations could be completed where the im-age data are stored, thus reducing memory access of image data and improving the performance and efficiency of image processing. Given consideration of image op-erations, the storage of image data in our architecture is different from the traditionalones, which should pay attention to both storage efficiency and computing efficien-cy. Compared with traditional system with hard disk, our architecture will reduce memory access overhead obviously, thus increasing the system performance.2. Designing the basic image operations and common basic logic elements in the ar-chitecture(Chapter 3)With the analysis of image processing algorithms, we abstract four types of basic image operations and their design in the proposed combining architecture. The four algorithms are image logic operations, threshold segmentation, edge detection and template operations. We design the detail of these algorithms in the above architecture according to the characteristics of memristor. In the designing process,we find some common used basic logic elements, including decoder, counter, shifter and adder, and design these basic elements in the proposed architecture. Compared with traditional design, the designs in this thesis are completed in memory, thus avoiding reading and writing of image data and improving the efficiency of the system.3. Validating the design of basic logic elements and analyzing the performance(Chap-ter 4)To validate the correctness of the architecture and design proposed in this thesis, we implement our design with experiments and analyze the performance. The exper-iments are simulating experiments with HSPICE software. First of all, we imple-ment the memristor model with HSPICE simulator, and investigate the influences of different parameters on the characteristics of memristor. We can choose proper parameters according to the request of the application. Next, we finish the simulation of basic logic operations with the memristor model, including a 3-8 decoder,4-bit counter, 8-bit shifter and adder, and validate their correctness. Then, we analyze the performance of basic logic operations and basic image operations, and compare them to the traditional image processing techniques. The results show that our design is correct and efficient.
Keywords/Search Tags:Image Processing, Memristor, Memristive Memory, Stateful Logic, Resistive Memory, Crossbar Array, Threshold Segmentation, Template Operation, Edge Detect
PDF Full Text Request
Related items