Seal removing method

Abstract

The invention discloses a seal removing method, and belongs to the field of computer vision. The method comprises the steps of providing a first original image with a seal area; positioning and segmenting a seal area from the first original image, and extracting a first seal image; performing normalization processing and scaling processing on the extracted first seal image, wherein the scaling ratio is Sh and Sw; adopting CycleGAN as a seal removal network, removing the seal in the first seal image, and converting the first seal image into a first seal-free image; amplifying the first stamper-free image according to a ratio of 1 / Sh to 1 / Sw; and restoring the amplified first seal-free image to an original image, and outputting a first seal-free original image. The method disclosed by the invention is simple in training and high in generation efficiency, and the algorithm has great robustness and great application value.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a stamp removal method. Background technique [0002] In various scenarios, documents need to be digitized, such as document image analysis, recognition, classification, storage, etc., all need to use the content information in the document. This recognition is called OCR technology. In applications such as bank form recognition, financial tax invoice recognition, etc., images are often stamped with picture (seal) seals, such as company official seals, company financial seals, personal name seals, etc., with round seals, oval seals, square seals, etc. Rectangular frame stamps, etc., stamp colors also come in various forms, most of them are red stamps, there are also blue stamps, grayscale scans are gray stamps, and black and white stamps after image binarization, etc. The stamp is placed on the paper, and in many cases, it covers the characters. Due to the influence of the stamp, ...

AI Technical Summary

Problems solved by technology

The stamp is placed on the paper, and in many cases it covers the characters. Due to the influence of the stamp, it affects the positioning and recognition of the text under the stamp, and indirectly affects the recognition rate of OCR.

[0003] The traditional method of stamp removal assumes that the stamp is red, and analyzes it according to the RGB value of the image to remove it. Due to the influence of various environmental factors, the color of the stamp is generally not pure red, some are purple, and some Black, etc., resulting in the removal of stamps based on the color separation scheme is not very clean, and even cannot be removed in many cases, especially for grayscale.

Images