An Approach for Improved Data matrix distortion correction using Median filtered Hough Transform

Data matrix is a two-dimensional barcode, widely used to store large amounts of information in a limited space. These data matrices are printed on labels or etched on surfaces where they are often subjected to various distortions and harsh conditions. This makes the data matrix unreliable and undetectable which calls for a need for effective and accurate detection. In this paper, an efficient image processing algorithm is presented to correct the distortion that utilizes a combination of median filter and Hough Line Transform to improve the accuracy of detection. Primarily, the raw image is preprocessed with a median filter to improve the quality of the hough line transform. The peak detected by the hough line transform corresponds to the angle of tilt necessary to be applied to the data matrix to correct the geometric distortion. The data matrix is then cropped and aligned according to the orientation of the finder patterns for further reconstruction. This implementation has been tested on a dataset containing 80 data matrices which have been curated by subjecting to various noises and distortion. Results indicate that the algorithm performs efficiently on the distorted images and produces accurate results with an error rate of 2.7% in the reconstructed modules of data matrices. Thus, the proposed framework can be implemented to recognize real-time distorted data matrix codes.