Abstract:

It is important to remove cloud in the CCD images of satellite for enhancing and extracting remote sensing information，especially in low latitude areas. In order to remove thin cloud in the CBERS-02B CCD images, two different forms of wavelet transforms were used. Low-frequency coefficients that are decomposed by two wavelet transforms are processed by homomorphic filter separately for lowing its low-frequency information. Results is reconstructed with low-frequency coefficients and high-frequency coefficients which decomposed from Mallat and   trous wavelet transform. According to quantitative evaluation,   trous wavelet transform combining with homomorphic filter algorithm can produce good results in thin cloud removal with higher entropy and standard deviation. It also can retain a large amount of information and rich detail information.

Key words: CBERS-02B, cloud removal, wavelet transforms, homomorphic filter