Research On Remotely Sensed Image Fusion Algorithms Of TM & SPOT

Multiple source remotely sensed images fusion technique has been concerned and applied widely in recent years. The theoretic and methods of it has became important research realm of subjects such as smart information processing, remote sensing application, etc. Multi-sensor image fusion is an important branch of multi-sensor information fusion, and is a technique to combine information from multiple source images of the same scene or object. Multiple source images of the same scene are usually from various sensors in the different time. By using multi-sensor, the fused image overcomes the limitation and otherness of single sensor image in geometry, spectrum, space resolution, etc. The experimental images of this paper are the multi-spectral image of Landsat TM and the high-resolution image of SPOT Panchromatic. SPOT panchromatic image has high resolution, and also has better special details than TM image. Otherwise, TM has low-resolution but better spectral information than panchromatic image. After fusing, we can get a new image with both multi-spectral and high resolution. And the result image we got has a more comprehensive and accurate description of the scene than single source image from TM or Pan. Now multi-sensor image fusion is widely applied in some fields, such as medicine, remote sensing, computer vision, weather forecast, automatic object detection, etc.In this dissertation, we focus on the improvement of algorithms of remotely sensed images fusion. The following main points are included:Firstly, we discussed the origin, evolution history and difficult problem of information fusion. And then we discussed levels, methods, fusing steps and evaluate standard of remotely sensed image.Secondly, we studied the pre-processing method of remotely sensed images.Thirdly, we studied the methods, realization and performance of popular fusion algorithms such as IHS transform, Brovey transform, PCA transform, etc. A fast-improved IHS transform was proposed and successfully used in image fusion experiment.Fourthly, A fast-improved IHS transform was proposed and successfully used in image fusion experiments.Fifthly, we studied the image fusion algorithms based on wavelet multi-resolution decomposition, compared the fused images with each other method. By comparison, results showed that IHS+DWT, PCA+DWT, WT based on feature, perform better than single DWT method.Sixthly, an improved fusion algorithm of remotely sensed images based on normalized correlation moment of characteristic values was proposed in this dissertation. We compared the fused image with the results gained by methods of traditional, normalized correlation moment and testified that the improved algorithm has better effects than all methods mentioned above.