This project develops computer vision methods for measuring the ruling of staves automatically in digitised music manuscripts, so that ruling patterns of pages can be compared. A stave’s five lines were ruled in one stroke with a rastrum, a pen with five fixed nibs, so staves ruled with the same rastrum show nearly the same spacing. Even though printed music paper emerged during the 16th century, most music paper continued to be ruled by hand. In the 18th century, often all staves were drawn at once, a type of music paper that is called machine ruled paper. Because notes and other marks cross the stave lines, the software does not trace lines one by one: it first estimates the layout of the whole page and then, guided by that layout, measures each line separately to a fraction of a pixel. For each stave it records the total span and the stave span, the distance between lines, the distance between staves and the length of stave. Accuracy is checked against line positions marked by hand.

Expressed as proportions of the stave’s height, the gaps are comparable across images of any resolution, even without a scale. A scale, given by hand or read automatically from a ruler photographed with the page, converts the gaps to millimetres, which separate rastra of equal proportions but different sizes. Line ends and distances between staves add clues, though these also reflect the pattern of the rastrum. Pages are grouped only if every pair agrees on each chosen measurement within a set margin. The result is a shortlist for checking by eye, not proof: two rastra can be made alike.

 

Hussein Mohammed and Oliver Huck
starting date: July 2026

Relevant projects: IRP 28, Wolfgang Amadeus Mozart’s ‘Fragments’