Research On Multifocus Image Fusion Based On Wavelet Decompositions And Structure Tensor Reconstructions

Image fusion as a subclass of information fusion, the corresponding process is to integrate the multiple images of the same scene obtained by different imaging equipments, or under different imaging conditions, or different imaging models, into a new image which contains the relative information of the source images. The propose of image fusion is to improve the reliability of the resulting image by processing the redundant data among source images, and to improve the definition and information capacity of the resulting image by processing the complementary information contained in source images. This fused single image may contain the more accurate and completed target information, is helpful for the perception of human vision and machine vision, and is suitable for the further image processing and analysis tasks. This technology has been widely used in many fields right now, such as military, medicine imaging, information security, remote sensing and industrial manufacture.The thesis deeply studies the image fusion methods based on wavelet decomposition in combination with structure tensor analysis. The main contributions of the research are as follows: 1. By systematically analyzing the traditional image fusion algorithms with wavelet decomposition, a selection method of low-frequency wavelet coefficients is provided based on the regional standard deviation and mean of the high frequency sub-band images in the same decomposition level.2. By introducing the structure tensor analysis method into the fusion process, a reconstruction algorithm of high-frequency wavelet coefficients in horizontal, vertical and diagonal orientation is proposed. The method can remain more details, enhance the contrast and get better fusion results in terms of both visual effects and quantitative evaluation.