Method and device for document image tilt detection

Abstract

The invention discloses a document image inclination detection method and device, and relates to the technical field of computers. A specific embodiment of the method comprises the steps of extractingimage feature points of a document image; Performing multiple rotation operations on the image feature points according to preset rotation granularity and rotation range, and determining a signal energy set; wherein the signal energy set comprises a signal energy value of each rotation operation; and screening out a maximum signal energy value from the signal energy set, and determining a rotation angle of a rotation operation corresponding to the maximum signal energy value, the rotation angle being a document image inclination angle. According to the method, the inclination angle is detected based on signal energy of image features, multiple document layouts are supported, the documents are not limited to be horizontal, vertical or mixed layouts, multiple image features are supported, the detection precision is high, and the calculation speed is increased.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a method and device for detecting the inclination of a document image. Background technique [0002] Slanted document image means that in the electronic process (scanning, etc.) of the document, due to the influence of external factors such as human beings, the scanned document is at a certain angle with the front edge of the image, that is, the phenomenon of tilting. This image is called a tilted document image. . Among them, document image tilt detection is a crucial step in the image preprocessing of OCR (Optical Character Recognition). When the document image has a serious text line skew, if it is not tilted and corrected, the detected text line may contain image information belonging to other text lines, which will lead to a decrease in the accuracy of the final OCR recognition result . However, skew detection has always been a very challenging problem, especially for d...

AI Technical Summary

Problems solved by technology

For example, the projection method and the Hough transform method are essentially based on image feature projection, calculation of variance, etc., and the calculation of this process is very time-consuming.

2. Algorithm detection accuracy is not high

3. The detection method has poor adaptability to different document layouts

For mixed layouts, this method is even more difficult to handle

4. The quality requirements for image feature extraction are too high

The existing traditional detection methods are all based on image features. When the image quality degrades, the image feature detection results will decline, which will eventually affect the accuracy of tilt detection results.

Images