Multicore Image Processing Platform And Its Application In Image Fusion Of Night Vision

With the development of image processing technologies, more advanced mathematical theory was introduced. At the same time, the mature application of imaging devices with high frame frequency and high resolution have also brought huge amounts of image data. However, for a long time, the contradiction between the data volume and complexity of the algorithm and its instantaneity has always been the challenge for image processing field. Considering the rise of multicore digital signal processor in recent years, this paper carries out the research on high performance image processing system based on multicore DSP structure.FMC6678 is a kind of image processing system with TMS320C6678, an eight-core DSP. On this basis, we implement a kind of image processing platform consists of one FPGA and two eight-core DSPs in this paper. Firstly, we bulid the framework of the whole system, then focus on the design of the multi-channel image acquisition, display, high-speed storage, high-speed transmission module and the corresponding driver for the FPGA image acquisition board. Besides, we implement three kinds of image preprocessing algorithm:median filtering, histogram equalization, image sharpening on it. Secondly, we study the gigabit Ethernet and SRIO interface between FPGA and DSP, then construct the high-speed image transmission mechanism between them. In addition, the transmission rate of the two kinds of high-speed interface are analyzed. Thirdly, we study the inter-processor communication method and multicore image processing framework of TMS320C6678, then analyze their respective applicable occasions.Finally, we study the Laplacian pyramid fusion algorithm and wavelet transform fusion algorithm of infrared and low level light image, then realize them in the parallel and pipeline processing framework respectively in the multicore image process system. Through experimental data analysis, we prove that the multi-core image processing platform can effectively increase the real-time processing capacity of image fusion algorithm.