The Study Of Multi-Source Remote Sensing Image Fusion Techniques At Pixel-Level

With the successful launch of a variety of remote sensing satellites and the rapid development of sensor technology, mass of remote sensing images with different spatial resolution, spectral resolution and temporal resolution are collected from different remote sensing platforms. Each type of remote sensing image has its own advantages and disadvantages, multi-source remote sensing image fusion can extract and integrate useful information from different types of remote sensing images of the same area and improve remote sensing data utilization. So, it has become a research focus in the field of remote sensing applications and image engineering. Multi-source remote sensing image fusion techniques at pixel-level are studied in this paper, the main tasks are:First, the common noise in remote sensing images and common denoising methods are elaborated and summarized in detail. Various denoising methods are used for simulation experiments to arrive at the most appropriate denoising method for each type of noise. For a mixed noise, which contains Gaussian noise and impulse noise, a combination algorithm of a multi-level median filtering and an improved wavelet-based threshold-denoising is used, experimental results show that the new algorithm is better than the traditional denoising methods.Second, subjective and objective evaluation methods of image fusion are introduced in detail from the concept, significance and application of principles. For the objective evaluation method, scholars have put forward a number of different indicators, which are fragmented and not very comprehensive. For this problem, common objective evaluation indicators are classified according to different purposes of image fusion.Finally, the weighted average fusion method, high pass filtering fusion method, Brovey transform fusion method, PCA transform fusion method, HSI transform fusion method and wavelet transform fusion method are specifically introduced in this paper. For how to gain appropriate weighted factor in high frequency coefficients fusion calculation of wavelet transform, wavelet transform fusion method based on the fuzzy sets is focused on, Gaussian membership function is used in the transformation of fuzzy matrix of high frequency coefficients. The results show that it is convenient to gain the weighted factor and the quality of fusion image is better. Because lαβcolor space transformation can effectively maintain spectral information, an improved image fusion algorithm which combines the advantages of wavelet transform fusion based on fuzzy sets and lαβcolor space transform is raised. Experiment results show that the improved image fusion algorithm is good.Meanwhile, on the basis of objective evaluation of fusion images, the above-mentioned fusion methods are compared according to different fusion purpose.