Research On Colorization Algorithm For Vehicle Infrared Video

Today, with the growth of automotive industry, there is growing emphasis on the function of the car, especially the security device is continuously applied to protect and guide the car. Many cars improve night vision technology in order to drive safely at night. The more traditional infrared images are rendered as gray images. However, the color level human eye can distinguish is hundreds of times than gray scale, so the development of an color infrared Automotive night vision system which can effectively improve driver night sight becomes an important issue at home and abroad. Through the research on Infrared video Colorization, this paper proposed several color night vision algorithms for the car.Firstly, according to car infrared image imaging principle, this paper describes system framework and characteristics of infrared image in detail. Based on the characteristics of infrared image we compare the latest algorithm to the classical algorithm and present a real-time color video with good color method vehicle for car infrared image.Secondly, in the phase of color video array, we apply motion estimation method to colorization for vehicle infrared video and propose colorization of night vision video based on motion estimation. After colorization of frames, we compare the brightness between of adjacent frames and tracking the movement of objects, thus rendering the next frame. The results also prove the feasibility and superiority of this method.Finally, a method for vehicle infrared video colorization is proposed which fuses multi-threshold image segmentation algorithm based on Fisher evaluation function and colorization algorithm based on prior color knowledge of infrared image according to the characteristics of infrared imaging. Firstly, we present fast fuzzy C means clustering segmentation on the key frame images to obtain cluster centers which set multi-threshold segmentation thresholds based on multi-threshold Fisher evaluation function. This can achieve a fast segmentation. Secondly, we classify the infrared video frame by frame using the above thresholds. Finally, we transfer color to the different classes in each frame of infrared video using priori color knowledge. Experiment shows that the algorithm of automatic colorizing vehicle infrared video is effective. These results indicate that the proposed method will benefit the target identification for human eyes and has a good real-time property.