| With the requirement for the real-time color synthesis of dual-spectrum night vision image in engineering application, this paper addresses this issue by focusing on the fusion and color transfer of the dual-spectrum night vision image.For the image fusion, through the analysis for the traditional fusion approaches,the multi-resolution analysis method is selected to realize the dual-spectrum night vision image fusion. The experimental analysis demonstrates that Nonsubsampled Contourlet Transform(NSCT) fusion method is most effective. In addition, because NSCT algorithm cannot manifest the useful information of the target effectively, a saliency method is introduced to boost the performance of image fusion. The result of experiment shows that this strategy not only can effectively fuse dual-spectrum night vision image, but also can highlight the useful information of the target robustly, which can obtain a better visual demonstration.With respect to the color transfer, conventional dual-spectrum night vision image color transfer may result in a much larger color deviation than actual expectation, this paper proposes two new color transfer methods. For the first one,traditional transfer methods in Lab color space are optimized and improved, where a stretch factor is introduced into Lab color space to color transfer, which can utilize the same reference image to conduct color transfer for the night images of different scenes. As for the second one, a method combined with a scaling factor and a steerable pyramid in YCb Cr color space is proposed. Experimental results show that the new strategy in YCb Cr color space is much better than the method in Lab color space and other classical color transfer methods.The proposed method for color synthesis of dual-spectrum night vision image is tested in DSP using C programming language. The result shows that the designed method can meet the requirements of real-time process in engineering application. |
PDF Full Text Request