A Novel Night Vision Image Color Fusion Method Based on Scene Recognition

Infrared and low light level image color fusion can make target detection and recognition more precise. Different from the existing color transfer fusion method using fixed reference image, this paper presents a color fusion method based on a combination of scene classification, fusion quality measure and color transfer. We introduce the scene classification method into the color fusion algorithm, which is based on the Gist descriptor and SVM classifier. Afterwards we use the proposed color fusion quality measure structure to find out the best matched reference image for each classified input image. Meanwhile, we get the high quality color fusion image using color transfer method. This method is verified in both linear and non-linear color space. Results show that this method can effectively improve the color fusion effect. More importantly, it can be used in the condition with few prior information.