Abstract: Aiming at the problem that digital images are vulnerable to attacks during network transmission and cause information leakage. This paper introduces quantum walk and Latin square matrix on the basis of Arnold scrambling, and designs a new type of color image encryption scheme. In the new scheme, the three channels of the color image are separated at first and the Arnold transform is used to scramble the pixel points of the image. Then quantum walk and Latin square matrix are used to process the pixel values of the scrambled image, the modulus diffusion algorithm is used on the processed image to further change the pixel value of the image. Finally, the color encrypted image is obtained by merging the three-channels encrypted images. Quantum walk, as an excellent random sequence generation tool, provides random sequences for algorithms. We conducted experimental simulations on the encryption algorithm, and analyzed the experimental results in the histogram, correlation, information entropy, noise attack, cropping attack and other aspects. The results show that the histogram of the encrypted image is distributed evenly, the correlation between pixels is close to 0, and the information entropy is 7.9993, close to 8, indicating that the algorithm has a relatively good ability to resist statistical analysis. After decrypting the encrypted image after noise attack and cropping attack, the original image information can still be seen, indicating that the robustness of the algorithm is good. The key space of the encryption algorithm is large enough and the key sensitivity is good. The NPCR(normalized pixel contrast ratio) value is close to 99.61%, and the UACI(unified average changing intensity) value is 33.45%. It has the ability to resist differential attacks.