Research Of Image Processing Method Based On The Music Score Scanning System

With the rapid development of computer technology, the steps of digital music is becoming more and more quickly, the traditional papery music input need manual work, but it will cost a lot of energy. This paper will introduce the music automatic scanning method based on image processing technology, and develop the music automatic scanning system.The music scanning system mainly includes two parts, image scanning hardware and image processing software. Firstly, scan the papery music. Using the front CCD camera scan the five line score,system records the scanning video, through machine vision libraries obtain scanning video, and preserves the video frame by frame, according to the length of the video calculate sampling frequency and finishes the video sampling work. Next proceed image pretreatment and image stitching steps. System implements Gauss denoising algorithm for the sampled music image, extracts the SIFT feature points of every image to get each features of the note, then by kd-tree matching algorithm matching feature points and RANSAC algorithm filtering feature points,according to the result calculate the mapping matrix, and use the method of image forward stitching to obtain the whole single line music image.Finally, remove the music lines and repair the notes. For realizing the music line detection, based on image neighborhood searching improved algorithm of hough line detection complete music line positioning to ensure the accuracy of notes sound. In order to solve the problem of interference of five lines to the notes segmentation, remove the lines, and repair notes by using FMM repaired algorithm, at last we can obtain the whole music image information.In this paper the innovative points are as follow:(1) According to the existing papery music input problem, this paper proposes the music automatic scanning method based on image processing technology, and develop the music automatic scanning system,realizes digitization of music information.(2) Based on SIFT feature points.kd-tree matching algorithm and RANSAC filtering algorithm, realize the automatic stitching of the music image, improve the accuracy of SIFT feature points matching between the adjacent music images.(3) This paper proposes based on image neighborhood searching improved algorithm of hough line detection, realize five lines positioning,and by using FMM repaired algorithm repair the neighborhood values of the damaged notes, obtain the whole music image information,effectively increase the segmentation accuracy, and improve the notes sound result.