According to the principle of the wavelet transform for removing noise in images and the glimmer-and-granule characteristics of low light level (LLL) images,the methods for choosing wavelet and decomposing levels when the wavelet transform is used to remove the noise in LLL images are analyzed. A conclusion that the noise removing requirement of LLL images can be met by using Haar wavelet and decomposing one level is reached.To choose the coefficient threshold for removing noise by wavelet transform,three histograms of wavelet coefficients are analyzed and the algorithm for threshold choosing is designed.On the basis of the designed algorithm and the characteristics of LLL images,an experiential threshold for removing noise is obtained.Through the simulation experiment and the time analysis of algorithm complexity,the trade-off between the real-time ability and the effectiveness of removing noise in LLL images is attained.