Research On Multi-Sensor Image Fusion Methods

In recent years, with the development of new imaging sensors and image processing technology, multi-sensor image fusion has received significant attention for both military and nonmilitary application, such as automated target recognition, battlefield surveillance, missile guidance, remote sensing,fire control and medical diagnosis. Image fusion can combine complementary and redundant information from multiple sensors to achieve improved accuracies and more comprehensive than could be achieved by the use of a single sensor alone, which benefits for the feature extraction and target recognition.The actual fusion process can take place at different levels (pixel-level, feature-level and decision-level) of information representation. This thesis focused on the algorithms of the pixel-level fusion process,including simple image fusion methods, multi-resolution image fusion methods and false color image fusion methods. This thesis presented several multi-resolution image fusion methods based on wavelet transform and multi-scale pyramid decomposition, such as the image fusion algorithm based on laplacian pyramid (LP),and the image fusion algorithm based on contrast pyramid (CP). In consideration of the characters of the infrared (IR) image, an algorithm based on adaptively weighted median pyramid decomposition was presented. At the same time, we also studied false-color image fusion methods and gave a biologically-based false-color image fusion method.Several objective and subjective quality assessments of fused image were given in this thesis. We compared the performance of various image fusion algorithms with these evaluation criteria. The influence on fusion performance of the number of decomposition levels, fusion rules and operators, the size of region applied to the multi-resolution image fusion method were studied in this paper. A series of experiments on image fusion were given in this article .The experimental results showed that the fusion methods and evaluation criteria are effectual.