An improved image encryption/decryption approach is proposed. In the encryption system, a binary amplitude image is encoded in the Fourier domain. Then, the encoded image is scrambled with a random binary phase image to produce the encrypted image. Both the encoding and encryption processes are done electronically. The encrypted image is stored in an encrypted phase mask of which the transmittance is proportional to the phase components of that image. In the decryption process, the encrypted phase mask is optically descrambled with another phase mask which has the transmittance identical to phase components of the random image used in the encryption process. The descrambled field is inversely Fourier transformed by a Fourier lens in order to reproduce the original image. The simplicity and the misalignment free characteristics are the most significant advantages of our method. The Fourier encoding process using an optimization algorithm with iteration technique is discussed in detail. The effect of noise on the quality of the decrypted image is investigated. The innovation of the system is demonstrated through simulations. The most critical issue of our technique is that the encryption process requires an optimization search which consumes some computation time so that the system is expected to be applied for unreal-time applications.