一种图像融合算法硬件实现

Hardware Implementation of an Image Fusion Method

研究了Laplace金字塔图像融合算法,针对红外图像和可见光图像采用传统图像融合算法融合后,融合图像细节和目标很难同时比较突出,提出了一种基于形态学的Laplace金字塔图像融合算法。该算法先将红外图像进行形态学处理,然后将形态学处理后的图像与Laplace金字塔融合图像进行二次融合获得最终融合图像。在以TI公司TMS320DM642芯片为核心的融合系统中,使用C语言实现此图像融合算法,并且通过优化代码,实现320×240大小的红外和可见光图像实时融合,算法处理一帧红外和可见光图像时间约为33 ms。通过实验证明,本文算法得到的融合图像细节和目标都比较突出,且能够在硬件中实现实时性。

Based on the study of Laplace pyramid image fusion method, a Laplace pyramid image fusion method based on morphology is proposed in order to solve the problem that details of visible light image and the target of infrared image hardly being prominent at the same time in the fusion image, which is got by the traditional image fusion method. The method first processes infrared image by morphological way, and then let the Laplace pyramid fusion image secondly fuse with the image morphologically processed to obtain the final fused image. In the fusion system based on TI＇s TMS320DM642 chip, this image fusion method was implemented by the C language and 320 × 240 infrared and visible light images real-time fusion image was got by code optimization, and the processing time per frame is about 33ms. The experiments show that the details and target are more prominent in the fusion image got by the proposed method, and which can be implemented in hardware real-time.