Due to the limitations of techniques, conditions, and methods for river ice field observation, it is difficult to obtain large-scale and long-sequence data regarding river ice processes. Remote sensing technology can be used to effectively collect and deal with vast amounts of complex temporal and spatial-scale information. In this study, a Landsat8 remote sensing image was used as the data source, and the optimal band combination was obtained through comparison of single-band information and joint analysis of the correlation coefficient, information entropy, and the optimal index. The maximum likelihood method of supervised classification was used to interpret and analyze the remote sensing image data from the Sanhuhekou River bend of the Inner Mongolia reach of the Yellow River over various periods in the winter of 2013 to 2014. Through analysis of the interpretation results, the growth and melting processes and characteristics of river ice over different periods in the study of the river bend were determined. The results of this study show that TM743 is the optimal band combination for interpretation of the remote sensing data, and the interpretation image can reflect the growth, melting, and transport processes of river ice at a large scale in the winter. In combination with meteorological and hydrological data, the interpretation image can explain typical river ice phenomena, providing technical support for analysis of river ice theory and mechanism studies.