Audio-visual combinations in multimedia contents give us more exciting experiences than only audio or visual information. To support audio-visual video creators, affective impressions of audio-visual contents have been explored to date. However, little research has analyzed in terms of structures or features of audio-visual contents. Revealing relationships between audio-visual impressions and their structures or features can be helpful for video creators. The current study aims to analyze effects of music on image impression in terms of musical features. Specifically, we asked participants to rate impressions for image stimuli at two conditions in which image-only stimuli or music-image combination stimuli were presented. Three factors representing image impressions were extracted by factor analysis, and differences of factor scores between two conditions were calculated to describe impression changes by music. We confirmed relationships between impression changes and musical features (e.g. volume, pitch, and timbre) by ridge regression analysis. As a result, regression analysis accounted for high variances in amount of impression changes and revealed that some musical features influenced on image impressions. This study provides a feasibility of estimating image impressions by musical features.