Hiding data in a medium such that it can not be perceived easily is called Steganography. Two of the most important performance metrics, of steganography are the embedding capacity and noticeability. Specific features of a cover image determine the amount of data to be embedded. A novel algorithm utilizing all pixels of a cover image and does not need any other information while decoding is proposed in this paper. The method computes intensity variance of 2x2 non overpapping subimages to decide the number of bits to use while encoding and decoding. Intensity variance of the subimage is computed after the dynamic range of the pixels are reduced by masking off increasing number of LSB's at every iteration. LSBs that result in less than 10 percent of variance change has been assumed to be insignificant in the variance computation and those bits are selected to embed data.