The quality of Sonar images are often reduced by the presence of speckle noise. The presence of speckle noise leads to incorrect analysis and has to be handled carefully. In this paper, an improved non-parametric statistical wavelet denoising method is presented. The algorithm uses a stationary wavelet transformation to derive the wavelet coefficients, from which edge and non-edge wavelet coefficients are identified. Further to improve the time complexity, only homogenous regions with respect to coefficients of neighbors are considered. This method uses an ant colony classification technique. A hybrid method that exploits both inter-scale and intra-scale dependencies between wavelet coefficients is also proposed. The experimental results show that the proposed method is efficient in terms of reduction in speckle noise and speed and can be efficiently used by various sonar imaging systems.