Visual Secret Sharing (VSS) is a method to encode a secret image into multiple shares. Coding one pixel of secret image into m sub-pixels in shares during encoding process causes pixel expansion problem. Probabilistic Visual Secret Sharing Scheme (PVSS) are proposed in recent years to overcome this problem. In traditional VSS and PVSS methods, human visual system is sufficient for recovering the secret image. In 2007, Wang proposed a (2, n) probabilistic method that uses Boolean operations for encoding and decoding of secret image. In this study, we proposed a method for generating random B matrices that are used by Wang's method. Accuracy of the reconstructed white pixels is improved. Maximum PSNR value of the reconstructed image with the proposed method will be higher than that of Wang's PSNR value. One of the most important criterion for evaluation of VSS schemes is accuracy of reconstructed pixels and proposed method yields improved accuracy compared to Wang's schemes.