Research For Dual-band Infrared Imagery And Date Fusion

Image fusion has been an important new technique for computer image vision.The essence of image fusion is to combine information collected from multipleimages of a same scene. We can draw either redundant or complement informationfrom multiple source images of various sensors. It is possible to improve thesignal-to-noise ratio by utilizing the redundant information to get more reliableresults. It is possible to obtain a composite image with more complete detailedcontent by fusing the complementary information. The fused image is describedmore completed than any individual source of all images using multisensory. Fusedimages are more useful for machine distinguish and human perception. It is also forfurther image processes such as feature extraction and object recognition. When oneimaging sensor may not provide enough information for recognition or sceneinterpretation, or in adverse environmental conditions, it is always possible to get animproved image by image fusion.The image fusion technique is discussed for target detection and human visionbased on application requirements of infrared target detection under complexbackground. The image characteristics of two bands of MWIR are analyzed andcompared. The image fusion algorithm processing is described and commonalgorithms are summarized after research on three levels of image fusion(pixel level,feature level, decision level). Weak infrared target detection has always been thefocus and difficulty of study. The way to draw the weak target from infrared imagesunder complex background is important. The performance and scope of the weakinfrared target detection methods existed are briefly analyzed based on study of thetime–spatial–spectral domain model of infrared images. Combined withdual-color infrared focal plane, a method based on the adaptive spectral domain isused to enhance SNR. A method of multi-windows is proposed to different noisepoints from the targets. The results can improve the detection probability and the response speed, and provides a new idea for infrared small target detection of asingle frame. A dual-band infrared image fusion algorithm based on human visualsystem is presented. The source images are decomposed by nonsubsampledcontourlet transform, and fusion rules based on region are used. It’s proved succeedby experiment results.