A method and device for image preservation applied to image rotation

Abstract

The embodiment of the present invention discloses an image preservation method and device applied to image rotation. First, determine the minimum circumscribed rectangle of the key information area after the sample image is rotated as the target circumscribed rectangle, and then determine the unknown pixel value area, the unknown The pixel value area is an area where there is no image display content in the image area before the rotation of the sample image after rotation, and the fidelity boundary is determined according to the positional relationship between the target circumscribing rectangle and the unknown pixel area, and finally according to the determined fidelity The image of the key information area after boundary interception and rotation is used as a fidelity sample image. The method and device intercept the region of unknown pixel values ​​introduced after rotation outside the fidelity image to obtain fidelity sample images, which can be continuously used to expand the number of training samples, enrich sample features, and improve the recognition and generalization of deep learning algorithm models ability.

Description

technical field [0001] The present invention relates to data enhancement technology, and more specifically, relates to an image preservation method and device applied to image rotation. Background technique [0002] Deep learning is a method based on representation learning of data in machine learning. At present, deep learning technology has been gradually applied to various fields and plays an important role in various industries. [0003] Applying deep learning to the field of OCR (Optical Character Recognition, Optical Character Recognition), detecting text areas in images and recognizing text content is also an application direction of deep learning at present. However, since there is no large-scale text detection and recognition data set available, and the cost of obtaining a large amount of labeled training data is high, it is necessary to use data enhancement techniques in the field of computer vision to rotate, translate, zoom, add noise, and blur images. Various ...

AI Technical Summary

Problems solved by technology

[0005] In view of this, the present invention provides a method and device for image preservation applied to image rotation, so as to overcome the problem of image distortion caused by unknown pixel values ​​introduced by image rotation in the prior art

Images