Abstract: Data hiding technology embeds additional data by modifying the digital media signal and does not affect the use value of the media itself. Reversible data hiding can not only hide additional data and extract additional data, but also recover the original image without distortion, so it is the research focuses currently. Because JPEG is the most widely used image format, reversible data hiding methods can be divided into two categories: the first one embeds additional data by modifying the DCT coefficients domain, which results in file-size expansion and visual quality distortion. The second one modifies the entropy coding domain, and the encrypted image generated by this method has no signal distortion compared with the original image, but the payload is limited, and the higher the payload is, the larger the file expansion is. To mitigate the problems existing in these two kinds of methods above, a multi-domain reversible data hiding algorithm for JPEG images is designed, which embeds the additional data by modifying the DCT coefficients domain and the entropy coding domain. Since two domains have different effects on visual quality distortion and file expansion, this paper focuses on distributing the payload reasonably. In this paper, the reason of file expansion is analyzed firstly when embedding the additional data in entropy coding domain, then a payload distribution algorithm based on VLC frequency histogram is designed to minimize file expansion and visual quality distortion. Experimental results demonstrate that the proposed algorithm significantly outperforms state-of-the-arts in file size expansion and visual quality.