Method for automatically registering scanned document images

Abstract

The invention discloses a method for automatically registering scanned document images, which comprises the following five steps: detection by using improved characteristic points, the coarse estimation of a rotation parameter, the coarse matching of characteristic points, the elimination of the erroneous matching characteristic points, and precise matching. Concretely, the five steps comprise: detecting the characteristic points of a standard scanned document image and a scanned document image to be matched by using the improved characteristic-point detection algorithm; estimating the primary rotation angle of the scanned document image to be matched relative to the standard scanned document image according to the positions of the characteristic points; calculating matched characteristic point pairs of the characteristic points in the standard scanned document image and the scanned document image subjected to primary rotation by using a normalized correlation algorithm; eliminating the erroneous matching by using an improved random sample consensus robust algorithm; and calculating the matching parameter between the standard scanned document image and the coarsely-rotated scanned document image to be matched according to the positions of the precise matching point pairs searched from an eight neighborhood, and making the parameter act on the coarsely-rotated scanned document image to be matched so as to obtain the registered image.

Description

technical field [0001] The invention belongs to the technical field of pattern recognition and information processing, relates to the processing of scanned images, and in particular provides a specific implementation scheme for automatic registration of scanned document images. Background technique [0002] Image matching refers to analyzing the characteristics of the overlapping parts of two images, and aligning the two images obtained under different circumstances. Usually, there are scale, rotation, and translation transformations between the two images. The matching process is to obtain parameters for these transformations. In people's daily work and life, scanners help people easily obtain images of documents. In most cases, it is necessary to realize automatic processing of documents, such as image segmentation, text extraction and recognition, etc., but the working process of scanners Rotation and translation (scales are negligible) are inevitably introduced in , whi...

AI Technical Summary

Problems solved by technology

[0004] The various methods proposed above are very complicated in terms of algorithm complexity, such as Hough transform, Fourier-Mellin transform, etc., which require image transformation, increase calculation consumption, cannot meet the application requirements in terms of speed, and do not make full use of scanned document images The characteristics of the scanned document image are that during the scanning process, the rotation angle and offset relative to the standard document image are relatively small, and compared to the standard document image, the images to be matched are all filled in on the basis of the former. Painted or annotated image

Images