:: Volume 7, Issue 2 (12-2017) ::
JGST 2017, 7(2): 17-26 Back to browse issues page
Introducing a Method for Detection of the Suspected Bit_Flip Noise Affected Pixels in Landsat Images
E. Amraei , M. R. Mobasheri
Abstract:   (788 Views)

Presence of different noises in Landsat images renders the extraction of descent information hard or sometimes impossible. One of these noises is Bit-Flip noise that happens when the signals are transferring from satellite to the other platforms and/or ground stations. During this transferring process, one bit might be affected and altered (e.g. 1 to 0 or 0 to 1). Clearly if the position of this bit is in higher bit values, then the change might be as large as the signal itself. This for dark pixels might be even few orders of magnitudes larger. In this work, a novel method for detection of affected bit and its correction is introduced. In this method, first we used a fuzzy detector method to identify the noise affected pixels. Then the pixels DN values were compared with the neighboring pixels and as results the affected pixels and the degree of affection were calculated. Since the change must be of the order of 2n, the proper value will be identified and corrected accordingly. The method was compared with the other well-known methods using statistical parameters such as SSIM and PSNR. The value for SSIM and PSNR were 0.9 and 28db respectively which were much better compared to other methods. 

Keywords: Bit-flip Noise, Satellite Images, Landsat, Remote Sensing
Full-Text [PDF 1289 kb]   (263 Downloads)    
Type of Study: Applicable | Subject: Photo&RS

XML   Persian Abstract   Print

Volume 7, Issue 2 (12-2017) Back to browse issues page