Method and apparatus for correcting oblique image

Abstract

The embodiment of the invention provides a method and apparatus for correcting an oblique image. Therefore, an oblique image can be corrected accurately with a low operation load and whether the corrected image is inverted is also determined; and if so, the inverted image is adjusted. The method comprises: cutting characters of an image; determining at least one feature character string from the cut characters; and correcting the image based on the at least one feature character string.

Description

technical field [0001] The invention relates to the field of computers, in particular to a method and device for correcting oblique images. Background technique [0002] Today, document information extraction has been widely used in various scenarios. The usual process of document information extraction is: first obtain the document image by scanning the document or taking a photo, and then use technical means to identify the key information in the image. However, in the process of taking pictures or scanning, due to various reasons, the document image may have a certain inclination, which will have a great impact on the extraction of document information in the later stage, which may easily lead to incomplete and wrong extraction of document information, or even failure to extract the document. information, so it is necessary to correct the skewed image. [0003] The process of oblique image correction in the prior art is generally as follows: [0004] S01. Using an imag...

AI Technical Summary

Problems solved by technology

[0010] 1. In the prior art, the Hough transform is used to calculate the tilt angle. Since the Hough transform needs to repeatedly calculate all the pixels, it requires a large amount of calculation and a long calculation time, which makes the image correction efficiency low

[0011] 2. The existing technology rotates by detecting the outline of the image. For an image deflected by 180 degrees, it is impossible to judge its front and back

[0012] 3. When affected by noise and the background and image information are not clear, it is difficult to extract the contour with the existing technology, resulting in the inability to achieve correction

[0013] 4. When using the Hough transform to detect straight lines in the prior art, due to the different backgrounds of the pictures, more interference lines are usually detected, so that the straight line of the image frame cannot be determined

Images