Study On Image Fusion Based On Wavelet

Image fusion is a field of Information Fusion in which research object is mainly image, which is comprehensive new technology including many objects, such as sensors technology, image processing, artificial intelligence, etc. Image fusion is the process by which multiple images of the same scene are combined to generate more complete and accurate description of the scene than any of the individual source images. Recently, the technology of image fusion has been applied to many fields, such as military application, remote sensing, medicine, computer vision, and so on.In this paper, we focus on the methods of pixel level image fusion. The algorithm based on transform domain is emphasized. We do profound and systematic research in multi-source images by multi-resolution analysis and put forward several fusion algorithms. The main research work and innovations are listed as follows:1. For multi-focus image fusion, algorithm of image fusion based on nonsubsampled Contourlet transform is proposed. According to imaging mechanism of multi-focus images and human visual system feature, fusion strategy of low frequency sub-band coefficients and band pass directional sub-band coefficients is proposed respectively in NSCT domain. For low frequency sub-band coefficients, fusion rules based on directional vector norm are proposed. After introduction of band pass directional contrast and taking into account the correlation between pixels, the improved band pass directional contrast is proposed. We propose fusion rules of band pass directional sub-band coefficients combing the improved band pass directional contrast and local area energy. This algorithm can preserve useful information from source images, eliminate artifact and suppress noise.2. For fusion of visible images and infrared images, algorithm of image fusion based on nonsubsampled Contourlet transform is proposed. According to different physical features of visible images and infrared images and human visual system feature, we present fusion strategy of coarse scale coefficients and fine scale coefficients in Curvelet domain. Modular Principle Component Analysis is introduced to the process of coarse scale coefficients fusion in order to preserve detail information from source images. Local area energy is used in the process of fine scale coefficients to extract object information from source images. This algorithm can preserve detail information of visible images and object feature of infrared images.3. Algorithm of visible images and infrared images fusion based on Curvelet transform and Wavelet transform is proposed combining the second generation Curvelet and Wavelet transform. Firstly, making most use of superiority of Curvelet transform and Wavelet transform in image processing, source images are decomposed to obtain coarse scale coefficients and fine scale coefficients by Curvelet transform. Fusion rules based on Wavelet transform are made out in Wavelet domain. For low frequency coefficients, fusion rules based on image physical feature are proposed; while for high frequency coefficients, fusion rules based on neighborhood directional contrast and local matching are considered. Fusion decision map is reconstructed to obtain fusion image by Wavelet transform and then Curvelet transform. This algorithm can preserve useful information from source images and strength object information.4. Algorithm of adaptive multi-focus image fusion based on Wavelet transform is proposed combining Wavelet transform and Pulse Coupled Neural Networks. Fusion rules are designed by making use of good features of Wavelet transform and PCNN in image processing field and combining imaging mechanism of multi-focus images. Fusion coefficients are reconstructed to obtain fusion image. This algorithm can preserve detail information from source images and improve definition of fusion image.