The invention discloses an adaptive image coding method based on compressive sensing. In the adaptive image coding method, the sparsity of various image blocks in a DCT (Discrete Cosine Transform) domain is judged, and then adaptive compressive sampling (CS) is carried out on the image blocks according to judgment results. By adopting the adaptive image coding method, in the case of low sampling rate, an image can be reconstructed with high quality and an accurate sampling proportion is provided, wherein, an average PSNR (Peak Signal to Noise Ratio) value of the reconstructed image can reach over 35dB when an average value of the sampling rate is 44%; and the PSNR values of all the image blocks of the reconstructed image are intensively distributed so as to obtain the optimum overall quality of the image.