Design And Optimization Of Multi - DSP Parallel System Image Fusion Algorithm

With the powerful data processing capability, multi-DSP parallel system plays a great advantage in the field of image fusion. Each of Infrared image and visible light image has its own advantages. To emphasize this issue, this paper elaborates image fusion of infrared and visible image based on multi-DSP parallel system. and tests the color consistency of multi-DSP parallel circuit board.Firstly, an image fusion system with the core of multi-DSP parallel processing is designed, image fusion for multi-DSP system’s overall structure, and each part of the system are introduced. Especially, the hardware and software design of data processing systems are described in details.Secondly, multi-resolution image fusion algorithm based on multi-DSP parallel system is discussed. On the basis of traditional Laplace fusion algorithm, an improved Laplace fusion algorithm based on regional energy and the rule of local area gradient is studied. Then, because of the limitations of the Laplace fusion algorithm, the BEMD fusion theory is discussed. In order to be implemented in hardware system, a FABEMD fusion algorithms is proposed, and the algorithms are verified by experiments.Thirdly, since the human eyes are more sensitive to color information, naturally colorization fusion algorithm is researched based on multi-DSP parallel system. In the YUV color space, the color feature of the reference image is transmitted to the source image, so the source image has similar color information to the reference image tone. In order to make the ultimate fusion image colors richer and brighter, This paper makes Laplace fused image as a source of information in the Y-channel, transfers color in the U、V channel, and get s the final naturally colorization fusion image which is suitable for the human eyes.Finally, in order to verify that the fused images output from the multi-DSP integration board are the same color, our group has developed a color conformance test system. First, the standard image is selected and abstracted, then The abstracted image is burned to the signal generator. At last the multi-DSP fusion board is tested on the lightness, hue, and total color difference through the color consistency test software.