Remote Sensing Image Enhancement Based On Fuzzy Algorithm

Proposed by E.L.Pruitt and developing into a very advanced detectiontechnology, REMOTE SENSING has been widely applied in the national economyand defense, creates enormous social and economy benefits and becomes a newmeans to understand the universe multi-dimensionally since1960s. In the process ofcollecting images, it is inevitably affected by various environmental factors, resultingin poor visual effects, low resolution, low brightness and other disadvantages. At thesame time, the collected grayscale range of remote sensing images couldn’t includethe ones of remote sensors. Therefore, the remote sensing image enhancementtechnology play a positive role in improve image contrast, highlight some localdetails.Image enhancement methods can be divided into two roughly: The method ofimage enhancement in spatial domain and the method of image enhancement intransform domain.The first method literally means enhance the image in the spatialdomain, in other words, it means process each pixel in the image directly. In this paper,we introduce the fuzzy algorithm similar to spatial processing method and someimprovements based on the traditional fuzzy algorithm.The second one means transform the original image to a specific domain andenhance the image by modifying the correlation coefficient in transform domain.There are some methods transing the Images from spatial domain to transform domain,such as Fourier transform, wavelet transform, Contourlet transform, NSCT transform,Shearlet transform.In this paper, we respectively apply the improved fuzzy algorithm to enhance theimage in spatial and transform domain, and propose two algorithms to enhance theremote sensing image:First, the improved fuzzy algorithm and a spatial image enhancement algorithmsare used to enhance the remote sensing image. Because of the traditional Pal-King algorithm is difficult to determine the threshold, we adopt the improved OTSUalgorithm to obtain threshold adaptively. Meanwhile, the enhanced details is notobvious due to fuzzy enhancement is globally processed, we adopt relative entropy asa judgement standard to enhance the neighborhood informational pixel. Thisalgorithm not only enhance the brightness of the original image but also the details ofimage. Experimental results show that we can get great visual effects and moreobvious details by this algorithm.Second, we adopt Shearlet transform to enhance the remote sensing image forthe first time, and propose a remote sensing image enhancement algorithm based onShearlet transform. After using fuzzy contrast enhancement and fuzzy enhancementon the low frequency coefficients and the high frequency coefficients of each scaleand direction, we get the final result by inverse Shearlet transform. Experimentalresults show that we not only get a good visual effects subjectively, but also have agood indicator and performance objectively by this algorithm.