Abstract: Aiming at the problem that the infrared image has the weakness of dim targets and poor details, is generally black-white image and not used for chroma subsampling, the paper introduces a new infrared imaging colorization algorithm based on local linear embedding（LLE）. The method seeks the mapping for the gray space to the color space, achieves the transfer from infrared image to colorized infrared image. Firstly this paper transfers the object infrared image and source color image to YUV color space, separates the luminance and hue information, secondly it makes each pixel and the neighbor pixels connected to vector, chooses equally part pixels of reference color image and makes them become vector using the same principle, then accomplish matching pixels between the object and reference images, and highlights the brightest parts of the resulted infrared image based on auto-threshold pseudocoding. Applying the method on infrared thermal imager of lab, it has color information that improves the visual effect, avoids visual fatigue, and makes the hot targets more noticeable using the sensitive color such as red and yellow, the experiment results show that it is advantageous to long-time observation and target distinguishing.