Research On Colorization Technology Of Night Vision Image Based On Color Transfer

In order to make the night vision image with a natural color, color transfer algorithm has been widely used. Color transfer is an emerging image processing method in recent years, which could transmit one image's color to another, and then,the two images have similar colors. In the color night vision technology, the night vision image will get the true color, and it is conducive for the observer to understand the scene and the target identification. It has a broad application prospects in military reconnaissance, security monitoring and so on. In order to get the night vision image with true color, several improved color transfer algorithms with good effect and faster computing efficiency are proposed in this paper.First of all, this paper researched on the pseudo-color algorithms and some classic color transfer algorithms. As one of the most widely used methods of color night vision, Pseudo-color algorithm is simple and fast, but the texture of the last color image usually can't be shown precisely. Therefore, color of the image obtained is not natural, and it does not meet the human visual habit. In order to obtain night vision image with natural color, color transfer algorithm has been widely used. Color transfer algorithm is divided into two categories: one is between color image and color image which used for image color correction, and the other is between color image and grayscale image, mainly used for grayscale images colorized.Secondly, we researched on the principle and steps of Welsh's color transfer algorithm, an improved color transfer algorithm based on Retinex image enhancement for night vision fused image is proposed against to the deficiency of Welsh's color transfer algorithm. First, the gray-level fusion of low-light level and infrared images was made with the pixel averaging fusion scheme, then brightness and contrast of the fusion image were enhanced using the proposed algorithm of improved multi-scale Retinex, and lastly used the improved algorithm of Welsh to complete color transfer. The results show that the matching accuracy was improved, and not only the fusion image's detail was significantly improved, but also the target is more noticeably.Finally, this paper proposed a color fusion algorithm of night vision image using color transfer and target enhancement. Frist, wavelet fusion and TNO method were employed to get better texture information of the false color image, then we select a similar reference image and make a conversion of the target image to reference image in the YCbCr color space, and match each of the channel's mean and standard to complete color transfer, with a contrast enhancement ratio in Cr channel which leads into outstand interest objectives to enhance the target. Experimental results show that:the algorithm not only maintain a higher efficiency, but also improved image detail,and the goal is more prominent.