In hyperspectral unmixing, endmember signals are not independent with each other, which compromise the application of independent component analysis (ICA) algorithm. This paper presented a novel approach based on constrained ICA for hyperspectral unmixing to overcome this problem. By introducing the constraints of abundance nonnegative and abundance sum-to-one, the purpose of our algorithm was not to find independent components as decomposition results anymore. In order to accord with the condition of hyperspectral imagery, we developed an abundance modeling technique to describe the statistical distribution of the data. The modeling approach is capable of self-adaptation, and can be applied to hyperspectral images with different characteristics. Experimental results on both simulated and real hyperspectral data demonstrated that the proposed approach can obtain more accurate results than the other state-of-the-art approaches. As an algorithm with no need of spectral prior knowledge, our method provided an effective technique for the blind unmixing of hyperspectral imagery.