In this paper an efficient (2, 2) secret sharing scheme has been proposed where the generated shares are meaningful and the aspect ratio and image dimension of the shares remain constant with respect to the source image. In the CARVCMS, instead of generating new pixels for shares, the shares are generated by random inversion of the levels of the pixel values. The scheme is more secured and very easy to implement like other existing techniques of visual cryptography. At the receiving end, the secret is revealed directly through human visual system by stacking the shares in arbitrary order with proper alignment. To expound the efficiency of the proposed technique, the results are compared with the technique developed by Naor and Shamir (1995), Hegde et al. (2008) and Jena and Jena (2009) where it has been shown that CARVCMS gives better performance. © 2011 IEEE.