Project Summary

This project aimed to detect and annotate musical notes on sheet music images using digital image processing techniques. Initially, using a mask on Fourier-transformed images and Hough transform identified staff lines, but the process was too slow. A more efficient approach involved creating a binary image from the input where pixels above a certain threshold were identified as staff lines. Template matching then detected noteheads by comparing each image segment to predefined templates using the correlation coefficient. Non-maximum suppression helped refine the detection by reducing overlapping bounding boxes. The staff lines were categorized into treble and bass clefs based on their position, facilitating note classification within defined regions and assigning notes to specific bins based on their vertical positioning. This process effectively identified notes, although some, especially those between staff lines, were challenging to annotate clearly.