GAP FILLING METHOD FOR ETM+-OFF IMAGES BASED ON THE WAVELET TRANSFORM

Abstract

An unexpected and undesired failure of the Scan Line Corrector in the ETM+ instrument on board of Landsat 7 caused gaps. Lost data are difficult, impossible, to restore, because of their uniqueness, which turned these images unusable in some regions of the Earth. Some research was dedicated to partially compensate this problem. The solutions can be single source, when non-gapped pixels are used to derive a new value, or multi-source, when the value is derived from another image, which is supposed to be very similar to the problem image. In this paper, a new multi-source gap filling technique, based on wavelet analysis for application to LANDSAT ETM+ imagery, is presented. First, the noisy ETM-off image is filled repeating the previous valid line. Then the wavelet transform is applied systematically to both images, the noisy ETM+-off image and an ancillary highly correlated image. As it happens in some wavelet-based fusion approaches, the representation of the ancillary image is substituted by the representation of the noisy image. Then the inverse transform is applied, using the wavelet coefficients that were computed for the noise free image, which produces a hybrid image. In the las step, the gaps of the ETM-off image are replaced by the values of the hybrid image. The method is illustrated with real Landsat data. The results show that the method is valid and able to combine information of the noisy image and the ancillary data.