Based on M-band wavelet transform and scale space theory,a feature points detection method for scale and rotation invariant image is presented.First,the image is transformed with a multi-scaled M-band wavelet.Then a second moment matrix is defined by computing the entropy of M-band wavelet coefficients.Finally,the feature points are detected by studying the eigenvalues of the matrix.Experiments prove that the proposed method performs better than state-of-the-art methods under image scaling and rotation.Moreover,it keeps robust under illumination change and noise disturbance.