KR100944272B1 - Apparatus and method for restoring bar code image - Google Patents

Abstract

The present invention relates to a bar code image reconstruction apparatus and method, and to propose a bar code image reconstruction apparatus and method for performing a non-uniform illumination prediction in combination with a bar code deblur when reconstructing a bar code image.

Bar code image restoration apparatus according to the present invention, the observation unit for observing the input bar code image; A modeling unit for modeling a predetermined barcode image by including a predetermined barcode image, a blurring kernel causing distortion in the predetermined barcode image, and respective functions representing non-uniform illuminance as components; And a reconstruction unit configured to restore the observed barcode image by optimizing the modeled barcode image to increase the similarity between the modeled barcode image and the observed barcode image.

Barcode image, non-uniform illumination, blurring, binary image

Description

Apparatus and method for restoring bar code image}

The present invention relates to an apparatus and method for restoring a barcode image, and more particularly, to an apparatus and method for restoring a barcode image accurately.

In order to decode the information of the barcode image, the edges of the bar and space patterns must be accurately restored and the gap thereof is measured. When reading barcode images with a video input device such as a CCD (Charged Coupled Device) camera, barcode images are blurred and noise is added due to the effects of ink smear, uneven lighting, and lens distortion. Mostly. Therefore, when the blurred image is used to restore the barcode image, false edges are detected and decoding fails. In order to solve this problem, various methods for deblocking barcode images have been studied.

An object of the present invention is to provide a barcode image restoration apparatus and method for effectively decoding a barcode image by reconstructing the barcode image in consideration of blurring and non-uniform illumination.

In order to achieve the above object, the present invention proposes an apparatus and method for reconstructing a barcode signal by estimating non-uniform illumination as well as a blurring kernel when modeling a blurred barcode image.

More specifically, according to an aspect of the present invention, the above object is an observation unit for observing the input bar code image; A modeling unit for modeling a predetermined barcode image by including a predetermined barcode image, a blurring kernel causing distortion in the predetermined barcode image, and respective functions representing non-uniform illuminance as components; And a reconstruction unit for reconstructing the observed barcode image by optimizing the modeled barcode image to increase the similarity between the modeled barcode image and the observed barcode image.

In addition, according to another aspect of the present invention, the above object is a predetermined barcode including a predetermined barcode image, a blurring kernel causing distortion to the predetermined barcode image and respective functions representing non-uniform illumination as components. Modeling the image; And restoring the observed barcode image by optimizing the modeled barcode image to increase the similarity between the modeled barcode image and the actually observed barcode image.

According to the present invention, it is possible to accurately reconstruct the barcode image by overcoming the problem that the error in the non-uniform illuminance affects the error in the deblurring by combining the deblurring and non-uniform illuminance prediction of the barcode image. have.

In addition, according to the present invention, when reconstructing the barcode image by combining the blurring kernel and the non-uniform illuminance, the barcode image can be more accurately reconstructed by utilizing the fact that the illuminance gradually changes in the spatial domain.

Hereinafter, with reference to the accompanying drawings will be described in detail a preferred embodiment of the present invention. In the following description of the present invention, if it is determined that detailed descriptions of related well-known functions or configurations may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. In addition, terms to be described below are terms defined in consideration of functions in the present invention, which may vary according to intention or custom of a user or an operator. Therefore, the definition should be made based on the contents throughout the specification.

1 is a diagram illustrating a barcode image under non-uniform illumination.

(A) and (b) of FIG. 1 show two barcode images obtained under non-uniform illumination. In FIG. 1A, the illuminance of the middle portion of the image is brighter than the outer periphery of the image. In FIG. 1B, the illuminance is brighter than the right side of the image. (C) and (d) of FIG. 1 show one-dimensional signals scanned from (a) and (b) of FIG. 1, respectively. Referring to (c) and (d) of FIG. 1, it can be seen that the barcode image is distorted due to non-uniform illumination.

However, conventionally, only a blurring kernel was considered when modeling a barcode image, and there is a problem in that the barcode restoration performance is degraded because the barcode image is restored based on such an incorrect model.

On the other hand, attempts have been made to improve the illumination device for the barcode scanner in order to alleviate distortion of the barcode image due to such non-uniform illumination, but non-uniform illumination cannot be completely removed. In addition, since a general-purpose handheld device is used as a bar code reader such as a mobile phone having a digital camera, it has been difficult to apply the improved lighting device to such a handheld device.

Therefore, the present invention improves the barcode image reconstruction performance based on the improved barcode image model by performing barcode image modeling in consideration of non-uniform illumination as well as a blurring kernel generally considered when modeling a barcode image. An apparatus and method are provided.

2 is a block diagram showing the configuration of a barcode restoration apparatus according to an embodiment of the present invention.

Bar code restoration apparatus according to an embodiment of the present invention includes an observation unit 210, the modeling unit 220 and the restoration unit 230.

The bar code image is input and observed by the observer 210. The modeling unit 220 models a predetermined barcode image to include a function of indicating a predetermined barcode image, a blurring kernel causing distortion in the predetermined barcode image, and non-uniform illuminance, respectively. The reconstructor 230 restores the barcode image by optimizing the modeled barcode image such that the similarity between the barcode image modeled by the modeling unit 220 and the barcode image observed by the observer 210 increases. The restored bar code image may be input to an information analyzer (not shown), and digital information may be interpreted and output.

The modeling unit 220 models the barcode by multiplying the blurred barcode image by the spatially varying illuminance field. According to an embodiment of the present invention, the barcode image is modeled by further considering additional noise and unknown constant background value. A barcode image modeled according to an embodiment of the present invention may be expressed as in Equation 1.

[Equation 1]

Here, t denotes a one-dimensional space position, and * denotes a convolution operator. I (t) represents an unknown illuminance, G (t) represents a blurring kernel, and x (t) represents a clean barcode signal.

According to an embodiment of the present invention, based on the assumption that the number K of edges in the barcode signal is known, the clean barcode signal may be parameterized as in Equation 2 below.

[Equation 2]

Here, e = [e ₁ , e ₂ , ... e _k ] represent positions of edges satisfying e _{i +1} > e _i , and u (t) is a unit step function.

In addition, a blurring kernel that causes blurring of a clean barcode signal may be parameterized by using a Gaussian function having an unknown standard deviation σ> 0 as shown in Equation 3 below.

[Equation 3]

Although the true blurring kernel is neither space invariant nor accurate Gaussian function, the spatial invariant Gaussian model has been effectively used to model the blurring kernel.

In addition, spatially varying illuminance, that is, non-uniform illuminance, may be modeled as in Equation 4 using a third-order B-spline function uniformly located in the spatial domain.

[Equation 4]

Where β ³ (t) is the third-order B-spline function centered at zero, L is the number of B-spline functions, and b = [b ₁ , ..., b _L ] is the B-spline function Coefficients, l _j , j = 1, ..., L represent uniformly distributed center positions of the B-spline function.

Modeling the barcode image, the non-uniform illuminance, and the blurring kernel with the functions as described above, respectively, is only an embodiment of the present invention, and the barcode image, the non-uniform illuminance, and the blurring kernel may be modeled in other forms.

Using the models in Equations 1 to 4, in accordance with an embodiment of the present invention, the objective of the combined performance of non-uniform illuminance prediction and deblur of a barcode image is obtained from y (t) to e , b , c. And accurately predict σ to restore the original barcode image. Accurate prediction of e is sufficient for accurate decoding of the barcode image, but accurate prediction of b , c, and σ is also necessary for accurate prediction of e . If the parameters b , c, and σ are not predicted correctly, then the prediction of e with inaccurately predicted parameters becomes equal to the prediction based on an incorrect signal model.

를 예측하는 문제가 된다. 여기에서, 파라미터 벡터의 개수 N은 σ＞0, e_{i +1}＞e_i, c＞0이고 b_j＞0와 같은 제약에 종속하는 관찰된 신호 y(t_i), i=1,...M의 샘플들로부터 N=K+2+L이 될 것이다. 이 문제를 해결하기 위한 효과적인 방법 중 하나는 관측된 신호와 제약조건을 만족하는 모델 사이의 제곱 에 러(square error)가 최소로 되는 파라미터 벡터 θ를 결정하는 것이다. This problem is a parametric vector for optimal design.

It is a problem to predict. Here, the number N of parameter vectors is the observed signal y (t _i ), i = 1,... Which is σ> 0, e _{i + 1} > e _i , c> 0 and is subject to constraints such as b _j > 0. N = K + 2 + L will be from samples of .M. One effective way to solve this problem is to determine a parameter vector θ that minimizes the square error between the observed signal and the model that satisfies the constraint.

That is, the restoration unit 230 restores the barcode image by optimizing the modeled barcode image such that the similarity between the barcode image modeled as described above and the barcode image observed by the observer 210 increases. To this end, the restoration unit 230 estimates the parameters e , b , c, and σ that minimize the square error of the modeled barcode image and the observed barcode image.

를 결정하는 것이다. As such, predicting the parameters e , b , c and sigma minimized by the squared error between the modeled barcode image and the observed barcode image becomes a constrained optimization problem. However, since most real barcode signals do not violate the constraints described above when the initial prediction is close to the optimal parameter vector, the problem is a non-constrained optimization problem. In the nonconstrained optimization framework, the most common method is a predictor that minimizes the sum of nonlinear squares.

To determine.

[Equation 5]

는 다음의 수학식 6과 같이 정의된다.Objective function here

Is defined as in Equation 6 below.

[Equation 6]

는 모델링된 바코드 영상과 관측된 바코드 영상의 차이로서 다음과 같이 정의된다 .From here,

Is the difference between the modeled barcode image and the observed barcode image.

[Equation 7]

는 표시의 단순화를 위한 것으로

을 나타낸다. 비선형 최소 제곱의 문제의 해를 구하기 위하여 가우스-뉴튼 방법이 이용될 수 있다. 이때, 대부분의 바코드 영상들에 대하여 위배되지 않는, σ＞0, e_{i +1}＞e_i, i=1, ...K, c＞0이고 b_j＞0, j=1,...L라는 제약하에서 비선형 최소 제곱의 문제의 해를 구할 수 있다. Where t _i represents the sample location,

Is to simplify the display

Indicates. The Gauss-Newton method can be used to solve the problem of nonlinear least squares. At this time, σ> 0, e _{i + 1} > e _i , i = 1, ... K, c> 0 and b _j > 0, j = 1, ... which do not violate most barcode images. Under the constraint of L, we can solve the problem of nonlinear least squares.

However, in Equation 5, the solution of the nonlinear least-squares problem may not be unique and thus may not be solved. In order to solve this problem, according to the present invention, the restoration unit 230 restores the observed barcode image based on the fact that the illuminance gradually changes in the spatial domain in order to optimize the modeled barcode image.

If this is expressed as an equation, it is to design a new objective function such as Equation 8 including a roughness penalty function.

[Equation 8]

는 다음과 같이 정의된다. Where λ> 0 is a normalization parameter and the roughness penalty function

Is defined as

[Equation 9]

Where R represents the matrix form of the roughness penalty function. The penalty function in Equation 9 allows the predicted roughness to be flattened by penalizing the change in neighboring B-spline coefficients. Where L is the distance between two neighboring B-spline coefficients. The first term in the objective function in Equation 8 causes the error between the observed barcode image and the modeled barcode image to be small, while the second term promotes flattening the predicted roughness. The normalization parameter [lambda] is a parameter that adjusts the ratio between the fidelity of the observed data and the flattening of the illuminance and can be determined manually.

On the other hand, accurate initial prediction is important for effective minimization of the objective function in Equation (8). According to one embodiment of the invention, the initial edge position e may be estimated using local maximum and minimum values of the signal obtained after wavelet-based denosing of the observed signal. In addition, σ is initialized using the distance from the last local maximum to the last local minimum in the spatial domain. The magnitude of the illuminance is estimated using the distance between the maximum and minimum values of the observed signal. Based on the assumption that the illuminance is near uniform, all B-spline coefficients are initialized by the predicted magnitude of illuminance.

If the actual illuminance is uniform and the actual sigma is very small, the initial illuminance will be predicted close to the actual illuminance. In addition, an initial value of a constant background value c may be obtained by dividing the minimum value of the observed signal by the predicted intensity of illumination.

In summary, the restoration unit 230 may include a function representing the barcode image of the modeled barcode image, a function representing the blurring kernel, and the non-uniform illuminance such that the similarity between the modeled barcode image and the observed barcode image increases. Estimate each feature value needed to determine the function represented. According to an embodiment of the present invention, a feature value of a predetermined barcode image is a value representing the position of an edge of the barcode image, and a feature value of the blurring kernel is a value representing a degree of distortion in the barcode image. (σ), and the feature value of the nonuniformity roughness is a coefficient b of a function representing the nonuniformity roughness. Then, the reconstructor 230 may reconstruct the barcode image based on each estimated feature value.

In addition, the reconstructor 230 estimates an initial value of each feature value from the barcode image observed by the observer 210 and then based on the estimated initial value of each feature value, the observed barcode image. Each feature value is estimated to have the highest similarity with.

3 is a flowchart illustrating a barcode restoration method according to an embodiment of the present invention.

A predetermined barcode image is modeled by including a predetermined barcode image, a blurring kernel causing distortion in the predetermined barcode image, and respective functions representing non-uniform illumination as components.

The barcode image is reconstructed by optimizing the modeled barcode image to increase the similarity between the modeled barcode image and the actually observed barcode image (S 320).

When reconstructing the barcode image (S 320), the optimization of the modeled barcode image to increase the similarity between the modeled barcode image and the actually observed barcode image may be performed based on the fact that the illuminance gradually changes in the spatial domain. .

In addition, the restoring of the barcode image (S 320) may include: a function representing the barcode image of the modeled barcode image and a function of indicating the blurring kernel such that the similarity between the modeled barcode image and the actually observed barcode image is increased. And estimating each feature value required to determine a function representing the non-uniform illuminance, and restoring a barcode image based on the estimated feature values.

According to an embodiment of the present invention, the feature value of a predetermined barcode image represents the position of the edge of the barcode image, the feature value of the blurring kernel represents the degree of distortion of the barcode image, and the feature value of the non-uniform illuminance is non- The coefficient of the function representing the uniform roughness.

Further, in order to estimate the respective feature values required to determine the function representing the barcode image, the function representing the blurring kernel, and the function representing the non-uniform illuminance of the modeled barcode image, The initial value is estimated. Based on the initial value of each estimated feature value, each feature value may be estimated to have the highest similarity with the actually observed barcode image.

4 is a diagram illustrating a barcode image restoration result according to a barcode image restoration method according to an embodiment of the present invention, compared to a conventional barcode image restoration method.

을 나타낸다. 도 4의 (b)는 균일 조도 모델을 가지고 비선형 최소 제곱 방법을 이용한 방법(NLS-UI)으로 도 1(c)에 도시된 실제 바 코드 신호에 대하여 예측된 패턴

을 나타낸다. 도 4의 (c)는 본 발명에 따른 비균일 조도 모델을 가지고 비선형 최소 제곱 방법으로 도 1(c)에 도시된 실제 바 코드 신호에 대하여 예측된 패턴

을 나타낸다. FIG. 4 (a) is an edge-based method of decoding a barcode signal using initial prediction of edges without considering the fact that the barcode image is a binary waveform and predicting the actual barcode code shown in FIG. Pattern

Indicates. Figure 4 (b) is a method (NLS-UI) using a non-linear least square method with a uniform illuminance model, the predicted pattern for the actual bar code signal shown in Figure 1 (c)

Indicates. Figure 4 (c) is a pattern predicted with respect to the actual bar code signal shown in Figure 1 (c) in a nonlinear least squares method with a non-uniform roughness model according to the present invention

Indicates.

(길이 2)는

(길이 1)보다 작다. 유사하게, NLS-UI 방법은 신호를 정확하게 복호화하지 못했다. 예를 들어,도 4의 (b)에서

(실제 바 코드에서 길이 1)은

및

(길이 2)보다 더 크다. 또한,

(길이 3)은

(길이 2)보다 더 작다. 기존의 방법들과 비교하면, 제안된 방법은 도 4의 (c)에 도시된 바와 같이 더 정확하게 패턴들의 길이들을 예측할 수 있었다. 바코드가 도 4의 (c)에 도시된 바와 같은 문턱값들을 이용하여 정확하게 예측될 수 있음은 명확하다. As shown in Fig. 4A, the conventional edge-based method did not correctly decode the bar code. For example, in the edge based method

(Length 2) is

Is less than (length 1) Similarly, the NLS-UI method did not correctly decode the signal. For example, in Figure 4 (b)

(Length 1 in the actual bar code) is

And

Is greater than (length 2). Also,

(Length 3) is

Smaller than (length 2) Compared with the existing methods, the proposed method could predict the lengths of patterns more accurately as shown in (c) of FIG. It is clear that the barcode can be accurately predicted using the thresholds as shown in Fig. 4C.

So far I looked at the center of the preferred embodiment for the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential characteristics of the present invention. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

1 is a diagram showing a bar code image under non-uniform illumination.

2 is a block diagram showing the configuration of a barcode restoration apparatus according to an embodiment of the present invention;

3 is a flowchart illustrating a barcode restoration method according to an embodiment of the present invention;

4 is a diagram illustrating a barcode image restoration result according to a barcode image restoration method according to an exemplary embodiment of the present invention as compared with a conventional barcode image restoration method.

Claims (12)

Observation unit for observing a barcode image; A modeling unit for modeling a predetermined barcode image by including a function indicating a predetermined barcode image to be observed, a function indicating a blurring kernel causing distortion in the predetermined barcode image, and a function indicating a non-uniform illuminance; And And a reconstruction unit for optimizing the modeled barcode image to increase the similarity between the modeled barcode image and the observed barcode image, and reconstructing the observed barcode image using the optimized barcode image. Bar code image restoration device. The method of claim 1, And the reconstructor restores the observed barcode image by optimizing the modeled barcode image based on the fact that the illuminance gradually changes in the spatial domain. The method of claim 1, The restoration unit, Estimating respective feature values required to determine a function representing the predetermined barcode image of the modeled barcode image, a function representing the blurring kernel, and a function representing the non-uniform illuminance, And reconstructing the observed barcode image based on the estimated respective feature values. The method of claim 3, Wherein each feature value includes a feature value of the predetermined barcode image, a feature value of the blurring kernel, and a feature value of the non-uniform illuminance, The feature value of the predetermined barcode image indicates the position of the edge of the predetermined barcode image, the feature value of the blurring kernel indicates the degree of distortion of the predetermined barcode image, and the feature value of the non-uniform illumination is Bar code image reconstruction apparatus, characterized in that the coefficient of the function indicating the non-uniform illumination. The method of claim 4, wherein The reconstructor estimates an initial value of each feature value of the predetermined barcode image, a feature value of the blurring kernel, and a feature value of the non-uniform illuminance from the observed barcode image, Estimating a feature value of the predetermined barcode image, a feature value of the blurring kernel, and a feature value of the non-uniformity roughness such that the similarity with the observed barcode image is the highest based on each estimated initial value. Bar code image restoration device characterized in that. The method of claim 1, The modeling unit further includes an unknown background value to model the predetermined bar code image, characterized in that the bar code image reconstruction device. Modeling a predetermined barcode image including a function representing a predetermined barcode image to be observed, a function representing a blurring kernel causing distortion to the predetermined barcode image, and a function representing a non-uniform illuminance; Observing a barcode image; And And optimizing the modeled barcode image to increase the similarity between the modeled barcode image and the observed barcode image, and reconstructing the observed barcode image by using the barcode image modeled to be optimized. Bar code image restoration method. The method of claim 7, wherein In the step of restoring the barcode image, Optimization of the modeled barcode image is performed based on the fact that the illuminance gradually changes in the spatial domain. The method of claim 7, wherein Restoring the barcode image, Estimating respective feature values required to determine a function representing the predetermined barcode image of the modeled barcode image, a function representing the blurring kernel, and a function representing the non-uniform illuminance; And And restoring the observed barcode image based on each of the estimated feature values. The method of claim 9, Wherein each feature value includes a feature value of the predetermined barcode image, a feature value of the blurring kernel, and a feature value of the non-uniform illuminance, The feature value of the predetermined barcode image indicates the position of the edge of the predetermined barcode image, the feature value of the blurring kernel indicates the degree of distortion of the predetermined barcode image, and the feature value of the non-uniform illumination is Bar code image reconstruction method characterized in that it represents the coefficient of the function representing the non-uniform illumination. The method of claim 10, Estimating each feature value required to determine a function representing the barcode image of the modeled barcode image, the function representing the blurring kernel and the function representing the non-uniform illuminance, Estimating an initial value of each feature value of the predetermined barcode image, a feature value of the blurring kernel, and a feature value of the non-uniformity illuminance from the observed barcode image; And Estimating a feature value of the predetermined barcode image, a feature value of the blurring kernel, and a feature value of the non-uniform illuminance such that the similarity with the actual observed barcode image is highest based on the estimated initial value; Bar code image restoration method comprising the. The method of claim 7, wherein The modeling bar code image is a bar code image reconstructing method characterized in that it is further modeled by including an unknown constant background value.

Images