Optimal affine image normalization approach for optical character recognition

Abstract

Optical character recognition (OCR) in images captured from arbitrary angles requires preliminary normalization, i.e. a geometric transformation resulting in an image as if it was captured at an angle suitable for OCR. In most cases, a surface containing characters can be considered flat, and a pinhole model can be adopted for a camera. Thus, in theory, the normalization should be projective. Usually, the camera optical axis is approximately perpendicular to the document surface, so the projective normalization can be replaced with an affine one without a significant loss of accuracy. An affine image transformation is performed significantly faster than a projective normalization, which is important for OCR on mobile devices. In this work, we propose a fast approach for image normalization. It utilizes an affine normalization instead of a projective one if there is no significant loss of accuracy. The approach is based on a proposed criterion for the normalization accuracy: root mean square (RMS) coordinate discrepancies over the region of interest (ROI). The problem of optimal affine normalization according to this criterion is considered. We have established that this unconstrained optimization is quadratic and can be reduced to a problem of fractional quadratic functions integration over the ROI. The latter was solved analytically in the case of OCR where the ROI consists of rectangles. The proposed approach is generalized for various cases when instead of the affine transform its special cases are used: scaling, translation, shearing, and their superposition, allowing the image normalization procedure to be further accelerated.

Print Character Recognition Method Based on Convolutional Neural Network

“…According to the traditional pattern recognition method, the original data must be preprocessed and manually selected and extracted before classification, and it is often the most difficult to determine the best classification features of the original data. If you ignore this step and directly input each pixel of the original image as a feature point, when the input image is too large, the data to be processed is unprecedentedly huge, which will lead to dimensional disaster, which is obviously not feasible in reality [9][10]. Moreover, various characters in natural scenes, especially HCs, have relatively large changes.…”

Invoice Recognition System based on Neural Network

“…Firstly, the invoice image in the dataset is gray-scaled, which preserves the brightness and contrast of the whole image on the basis of retaining the current image information [13].…”

Deep Learning in Intelligent Navigation

“…The complex Network structure formed by the sequential linking of multiple Artificial neurons that can realize specific prediction function is called Artificial Neural Network (ANN) [13]. The core idea of artificial neural network is to set and organize a linear model with nonlinear activation functions (such as Sigmoid, TANh, Relu function, etc.)…”