A Study On The Adaptive Image Fusion Algorithm For Infrared And Visible Images In Different Illuminating Conditions

With the development of digital image processing technology, image fusion technology has attracted expanding concerns from researchers. In recent years, a variety of image fusion algorithms greatly advanced the development in this field. Due to its unique merit in providing complementary information, the image fusion of infrared and visible images has been applied to many fields such as military technology, remote sensing, medical, and targeting. Visible image falls to the category of reflected image, which is literally the kind that is affected by the surrounding environment. Infrared images are thermal radiation images, which mainly reflect the temperature difference between the measured object and the surrounding environment. In different lighting conditions, one camera might acquire dramatically different images. This paper is intended to design an adaptive image fusion algorithm using wavelet transformation, which has the ability to incorporate the images which are affected by different lighting conditions.This paper first introduces the three categories of image fusion algorithms, with main focus on the several most widely used pixel-level image fusion algorithms. To separate light images based on the light intensity, this thesis designs a BP neural network classifier. An adaptive wavelet transformation image fusion algorithm is also proposed: The visible light images are classified by the BP neural network and the neural network’s predicted outputs are combined with the weight age of the image fusion algorithm. This combination can achieve adaptive fusion effect by changing different fusion rules. Due to the fact that this algorithm is combined with a BP neural network classifier, this image fusion algorithm has the ability of self-learning. Experiments have been designed to compare this algorithm proposed here with other image fusion algorithms. This thesis also describes the quality assessment criteria for image fusion algorithms, which can evaluate the performance of the fusion algorithm proposed herein.