A noise-enhanced contrast stretching algorithm for enhancement of dark images in SVD-DWT domain has been presented in this paper. A dark or low-contrast image is considered to be comprising a weak signal (information) and noise (due to insufficient illumination). Since singular values of an image represent luminance of independent image layers, the internal noise may be considered to be inherent in the singular values of the image, as well that of the wavelet approximation coefficients of a dark image. The singular values of the approximation coefficients are iteratively processed using the analogy of a double-well system exhibiting dynamic stochastic resonance (SR), where addition of noise is utilized to enhance system performance. Iteration is terminated with the criterion of target contrast and visual qualities. Comparison with other SR-based and non-SR based algorithms shows noteworthy performance of the algorithm in the SVD-DWT domain. © 2013 IEEE.