Design And Implementation Of FPGA And Multi-core DSP Image Processing System

Digital image processing technology,is occuping a really significant position among all the advanced information processing technologies nowadays.Following the rapidly developing of ultra-large-scale integrated circuit manufacturing process,processors can be embeded with large amount of computing resources.Processors with high processing power,make the real-time processing of high data-rate digital image data possible.Real-time image processing technologies,especially the technologies based on multi-core DSP+FPGA architecture,have focus the attention of related experts all over the world.In this paper,a real-time image processing system based on XILINX V6 series FPGA and TI TMS320C6657 dual-core DSP is designed and implemented.The system select the advanced multicore DSP as the main processor of image data operation.DSP realizes the core algorithm of digital image real-time processing.The system uses Virtex6-series FPGA supplied by XILINX as a coprocessor of the DSP.Besides,FPGA also functioned as the main control unit of the whole system.The digital image input front end of the system is composed by SDI digital input terminal,USB3.0 transmission module and PAL video image acquisition module.Multi-core DSP and FPGA can transmit digital image data between each other duplexedly through the u PP transmission mode or Rapid IO high-speed transmit channel.Real-time image processing results can not only be transmitted to DVI,PAL display module,but also be send back to the host PC through the USB3.0 transfer module.In the process of system implementation,the author adopts USB3.0 transmission module as the digital image data input front end of the system,and realizes a high-speed,hot-swappable duplex data communication mode between real-time image processing system and PC host.The FPGA receive the high speed digital image data stream transfrom the host PC through the USB3.0 interface.Then the FPGA send the cached data stream to the multi-core DSP via the u PP peripheral interface.After processing the parallel image data stream,the multi-core DSP transmit the processed result back to the FPGA through the u PP interface.Finally,the FPGA transfer the processed digital image data to the display functional module,or send the data back to the host PC through the USB3.0 receive module.In addition,in order to cooperate with the multi-core DSP,make the system achieve better accuracy and velocity in real-time image object tracking and recognition.The author completes the theoretical research and real-time implementation of the image shape feature extraction algorithm on the FPGA embeded on the self-designed image processing platform.The FPGA divides the 8-bit grayscale digital image,then extracts the important image shape features such as the edge of the image,the connected domain of the image components,the centroids of the abstracted connected domains.These parameters are extremely important for the following image tracking and matching operations implemented in the DSP.By offering the reference data above,The FPGA can improve the stability and operational efficiency of the overall image processing system greatly.