Color night vision research based on multi-resolution color transfer

In order to acquire color night vision images accordant with human perception, and help observers to get more abundant scene information, a kind of color night vision method based on color transfer in multi-decomposition structure is proposed. Concerning their characteristics, low-level light visible and IR images are combined into a color fused image (source image) based on opponent vision property in YUV color space. Then the steerable pyramid is used to decompose the fused image and a colorful day-time reference image into multi-scale sub-band images. The mean and standard deviation of each sub-band image are calculated. Each sub-band coefficient of source image is modified according to the deviation ratio of the reference image to the source image in respective color channels. Finally, reconstruct the transferred image and obtain a low-level light visible/IR color night vision image with comfortable color appearance similar to the reference image. The transferred results in lαβ, YUV and RGB space are analyzed. YUV space is recommended to be the best color space for multi-decomposition color transfer. Comparing with traditional linear color transfer, the proposed method can make a color night vision image more realistic and natural, enhance the details, improve the situation perception and target detection ability.