A study on image-based music generation

Visual and auditory forms have noticeable associations that can inspire similar cognitive and aesthetical experiences. This study investigates the possibility of applying various visual-auditory associations in music generation. Several algorithms are proposed in order to create music segments from image features such as contour, colour and texture. Test results show that the algorithms can produce interesting and pleasant music segments in many cases. Integrating the generated segments together can create longer and more intriguing results. The generation process totally depends on the image content and avoids the problem of introducing pseudo randomness in many algorithmic music composers. Different images with similar features are able to generate similar music patterns. For very different images, some of the algorithms can generate music of different styles. This study shows that using image as a source of music generation has a high potential in creating interesting and fresh music.