Hyperspectral image (HSI) with high spectral resolution often suffers from low spatial resolution owing to the limitations of imaging sensors. Image fusion is an effective and economical way to enhance the spatial resolution of HSI, which combines HSI with higher spatial resolution multispectral image (MSI) of the same scenario. In the past years, many HSI and MSI fusion algorithms are introduced to obtain high-resolution HSI. However, it lacks a full-scale review for the newly proposed HSI and MSI fusion approaches. To tackle this problem,this work gives a comprehensive review and new guidelines for HSI-MSI fusion. According to the characteristics of HSI-MSI fusion methods, they are categorized as four categories, including pan-sharpening based approaches, matrix factorization based approaches, tensor representation based approaches, and deep convolution neural network based approaches. We make a detailed introduction, discussions, and comparison for the fusion methods in each category. Additionally, the existing challenges and possible future directions for the HSI-MSI fusion are presented.