A novel adaptive watermarking algorithm in discrete wavelet transform (DWT) based on quantization index modulation (QIM) technique is presented. The host image is decomposed into wavelet subbands, and then the approximation subband is divided into non-overlapping small embedding blocks. The secret watermark bit is embedded into singular value vector of each embedding block by applying QIM. To improve the invisibility and robustness of watermarking system, the quantization step for each embedding block is set by combining statistical model with particle swarm optimization (PSO) algorithm. The experimental results show that the proposed algorithm not only preserves the high perceptual quality, but also effectively stands against joint photographic experts group (JPEG) compression, low-pass filtering, noise addition, scaling, and cropping attacks, etc. The comparison analysis demonstrates that our scheme has better performance than the previously reported watermarking algorithms.