To overcome the negative effect of factors such as illumination and expression on face recognition, an adaptive feature and weight selection method was proposed. The method was based on Gabor image for face recognition. Firstly, 40 independent feature matrices which were reconstructed with the same scale and the same direction transform results of the different face images were obtained by regarding every Gabor wavelet transformed output image as an independent sample. In order to enhance the robustness to facial expression and illumination variations, the contribution of each new feature matrix could be adaptively computed by the proposed adaptive weight method. Secondly, after applying discrete cosine transform to each feature matrix, the coefficients which had more power to discriminate different classes than others were selected by discrimination power analysis to construct feature vectors. And, linear discriminant analysis features were extracted to fulfill recognition task. Experiments on the face databases demonstrate the effectiveness of the proposed method.