Colorization of short wave infrared image based on depth learning

Short-wave infrared imaging has the advantages of clear target detail expression, strong target identification ability, strong adaptability to haze climatic conditions and dust and smoke application environment, and can realize hidden imaging, low light level night vision detection, so more and more short-wave infrared imaging research is carried out. Based on the short wave infrared images as the goal, to carry out the short-wave infrared image colorization method based on color migration study, put forward an improved way of automatic unsupervised grayscale image coloring method, based on the depth of the convolutional neural network learning coloring method, and the global image of the transcendental and extracted from the whole image to calculate the local image features from image patches, The global prior provides image-level information, while the local feature represents the local texture or object at a given location. By combining these two features, the image is automatically colored to realize the color of short-wave infrared image, and the image quality evaluation after color is given. The experimental results show that the method has no loss of color image details, the image color information distribution is reasonable, high definition, obtained a new visual effect image, for target detection and recognition, key area monitoring to provide a more visual effect of target detection and recognition and scene monitoring means.