In the signal sparse representation methods, the traditional sparse approximation based on transform basis cannot extract the texture features of image adaptively, and the sparse approximation algorithm based on the over-complete dictionary is too complex. To solve this problem, a sparse representation method based on sparse dictionary optimization of wavelet transform is proposed. Based on the wavelet transform of image, this algorithm constructs over-complete dictionaries, and makes use of the similar attributes in interior and exterior on the texture in wavelet transform of images in the same scene, and classifies the over-complete dictionary with grey correlation degree, which improves the effectiveness of sparse representation. The new algorithm is applied in image signal sparse representation and image sampling and reconstruction experiments based on compressive sensing theory. The results show that the new algorithm improves the peak signal to noise ratio (PSNR) and structural similarity (SSIM) of reconstructed images as a whole, and can shorten the time of image reconstruction effectively.