【作者】 朱童； 张立保；

【Author】 ZHU Tong ZHANG Li-bao,(College of Information Science and Technology,Beijing Normal University,Beijing 100875,China)

【机构】 北京师范大学信息科学与技术学院；

【摘要】 提出一种基于平移不变与递推GCV(Translation Invariant and Iterative Generalized Cross Validation,TI-IGCV)算法,能快速获取小波去噪中的逼近最优阈值,并抑制奇异处产生的伪吉布斯(pseudo Gibbs)现象。通过提升小波对遥感图像进行多分辨率分析,利用GCV函数内部相关性并结合小波系数幅值分布,对各相邻整数阈值下GCV函数值进行快速递推计算,以满足遥感图像信号去噪的实时性需求。本算法还进一步借助平移不变实现更优的去噪效果且保持信号的平滑。在多组遥感图像去噪处理中,TI-IGCV算法耗时仅为GCV算法的6.5%至11.7%,TI-IGCV算法去噪后图像的PSNR比GCV算法去噪后图像的PSNR高出0.6-0.9db,表明本算法在遥感图像实时处理的效果和时间上均具备明显优势。更多还原

【Abstract】 This paper presents a fast and effective TI-IGCV (Translation Invariant Iterative generalized cross validation) algorithm for remote sensing image de-noising,based on integer wavelet transform.Distribution of Integer wavelet coefficients is calculated,followed by the join operation.Then the Iterative calculation of GCV functions is applied,regarding the relevance between GCV functions for adjacent thresholds.The proposed algorithm further improve the denoising result by employing cycle-spinning to reduce Gibbs phenomena near the discontinuities.The experimental result demonstrates that the time complexity of IR-GCV is 90% lower than that of GCV,while the PSNR of denosing image is 0.6db-0.9db higher.更多还原