Although the general random scrambling based on pixel can achieve a good chaotic effect, but it can not change the histogram of a digital image. We introduce the random scrambling into the domain of discrete cosine transform (DCT) of image and scramble the coefficients of DCT to improve the performance of scrambling. Firstly, we did 2-D discrete cosine transform to the original image. Secondly, we scanned the coefficients matrix of DCT by Zig-Zag scanning to get a 1-D sequence, and then we scrambled this sequence by 1-D random scrambling algorithm. Thirdly, we did inverse Zig-Zag scanning to the scrambled sequence and reconstructed the scrambled image from the chaotic coefficients matrix by 2-D inverse discrete cosine transform. Finally, to further improve the scrambling degree of the reconstructed result, we scrambled it again in space domain to gain the encryption image. Experiments show that this algorithm is effective at visual evaluation and is more stable in scrambling degree than Arnold transformation.
