An image classification method for tabular documents based on frame features and pixel distribution

Abstract

The invention relates to a table document image classification method based on frame line features and pixel distribution. First, grayscale and binarization are performed on the image to be classified, and then a frame line enhancement operation based on connected domain analysis is performed on the obtained binary image. Perform image denoising based on the deep learning method on the image after frame line enhancement; for the preprocessed image to be tested, use the morphological-based line detection method to detect and extract the horizontal and vertical frame lines respectively, and refine them , and then use the Npcanny-based line detection method to obtain the frame line number information; project the horizontal frame line image in the horizontal direction, and perform the vertical direction projection on the vertical frame line, and record the projected position and pixel value. The length information is matched with the standard template information that has been entered in the template library, the template image with the highest similarity with the image to be classified is selected, and finally the classification result of the image to be classified is output. The present invention can effectively classify table document images.

Description

technical field [0001] The invention relates to the fields of morphology and computer vision, in particular to a table document image classification method based on frame line features and pixel distribution. Background technique [0002] With the continuous development and progress of the economy and society, various industries and departments in the country will generate a large number of table documents in daily production and life. Manually classifying these table documents is not only inefficient, but also the classification accuracy cannot be guaranteed. The classification features of table documents generally include titles, frame lines, special characters, etc. Since the method for classifying table documents based on frame lines is more versatile, the frame line features are used here as the classification features of table documents. [0003] The table document image classification detects and extracts the frame line features of the table document image read into t...

AI Technical Summary

Problems solved by technology

[0002] With the continuous development and progress of the economy and society, various industries and departments of the country will produce a large number of form documents in daily production and life. Manually classifying these form documents is not only inefficient, but also the accuracy of the classification cannot be guaranteed.

Images