Abstarct: The emerging of the theory of compressed sensing (CS) is to be a promosing way to develop effective and new compression techniques, although the main concern is to reducing the sampling rate to reduce sampling costs. In this thesis, we use a CS according to block-baxsed coding that combine the discrete cosine transform and compressed sensing theory . Also we use two effective techniques to improve the reconstructed image. One of these techniques called coefficient random permutation(CRP) and the other one is measurement matrix weighting. CRP method effectively balances the sparse sampling of the vectors used in the DCT domain. This method improves the sampling with CS and can be used to encrypte the input image. The purpose of the weighting matrix measurement method is designing of adaptive measurement matrix. This design baxsed on energy distribution characteristics in the DCT domain. This method has a positive impact on the strengthening of the reconstruction of CS. Simulation results show that our proposed method reduces the image size by using block-baxsed CS and effectively improves the quality of reconstructed image. Image reconstruction baxsed on CS proposal according block-baxsed can be used for encrypted image compression and robust image compression application.