Strip Noise Detecting And Removal Of GF-1 Satellite Images

Known as China's first high-resolution earth observation system of national science and technology major project satellite,GF-1 satellite has been widely used in land and resources management,environmental monitoring and forecasting,modern agricultural monitoring,disaster prevention and mitigation,public safety and other major areas.Due to the long running time of the satellites in the trajectory,the GF-1 satellite remote sensing images are disturbed and influenced by various noise sources during the generation and transmission period,hence many strip noises exist on the satellite remote sensing images.The strip noises on GF-1 satellite images have seriously affected the interpretation and information extraction of image data,which heavily reduce the usefulness of the data.Many scholars have proposed numerous algorithms for strip noise removal of satellite images,which have a certain removal effect on specific types of strip noise or periodically distributed strip noise.However,most of the algorithms deal with the entire image,which will inevitably lead to the changes of pixel gray value on the non-strip noise image area.Motivated by the above considerations,a multi-type strip noise removal algorithm based on the strip noise on GF-1 satellite images is proposed in this paper.The proposed method is a multi-method combined method,which considers the cloud and water detection results.Finally,we make a qualitative and quantitative comparison of the results before and after the process of strip noise removal.The main contents of this paper are summarized as follows:(1)A strip noise removal algorithm which considers the results of cloud and water detection is proposed.During the process of strip noises removal in this paper,the image is first subjected to cloud and water detection,and then according to the results of cloud and water detection,the location of the strip noise is detected based on the statistics of the gray information.(2)A variety of detection methods are proposed to detect different types of strip noise on GF-1 satellite images respectively,which can detect different types of strip noise effectively.(3)Some proposed methods are capable of repairing different types of strip noises.The left and right neighborhood interpolation method is used to remove the single column of strip noise which has been detected.Multiple columns strip noise is repaired by the improved moment match method,which is based on calculating the reference mean and reference standard deviation by the weight of distance.For the multi-column strip noise,an improved moment match which adopts the linear fitting to calculate the reference mean and reference standard deviation is proposed.(4)Using the proposed method in this paper,the experiment of GF-1 satellite image with cloud and water area is carried out.The effects of strip noise removal and image restoration are significant.Besides,we take both the qualitative and qualitative evaluation into account to evaluate the repairing results.