CN105898104A - Method for substracting backside content revealed from scanned image - Google Patents

Abstract

The invention relates to a method for substracting backside content revealed from scanned images. The method comprises 6 steps of: (1) scanning a page padded with a white underlay and a black underlay at the back respectively to obtain 2 scanned images Iw and Ib; (2) utilizing an image key point detection and description method for detection to obtain key points of the images Iw and Ib and descriptors thereof, and matching the descriptors with the two groups of key points; (3) determining a rigid body transformation parameter between the images Iw and Ib by utilizing the matched key points; (4) selecting and marking first several pixels with highest numerical values of brightness gradient modulus on the image Iw; (5) establishing linear mapping fI of pixel brightness values between the images Iw and Ib; (6) and with regard to each pixel on the image Iw, replacing a brightness value thereof with a fI inverse transformation value of a pixel value of a corresponding point on the image Ib if the brightness value is less than a fI inverse transformation value of a brightness value of the corresponding point, otherwise, remaining the brightness value unchanged.

Description

A kind of method cutting down the back side content appeared on scanogram

Technical field

The present invention relates to image scanning and image processing field, a kind of abatement is appeared on scanogram The method of back side content.

Background technology

Image reading apparatus is that one is converted into digitized map the graph-text content on the page (one pages in such as books) The instrument of picture.

There is graph-text content on the two sides of some page, with image reading apparatus scan wherein one side on graph-text content Time, the content at the back side may appear, and printing opacity got over by paper, appears phenomenon the most serious.The back side content appeared The quality of scanogram can be reduced.

The method of the page back side content appeared in existing abatement scanogram has two kinds.One is at the page behind Cut down plus black substrate appear back side content (United States Patent (USP), 1991, the patent No.: 5053818, Title: " Method and apparatus for copying semitransparent originals using A black backing member "), the major defect of this kind of method be can appear material texture, defect, And the dark border of page outer ring.Another method is that first scanning obtains front digital picture and back side number Word image, then cuts down back side content (international appeared on direct picture by the method for image procossing Art meeting paper, 2009, author: Anna Tonazzini, Gianfranco Bianco, and Emanuele Salerno, exercise question: " Registration and enhancement of double-sided degraded Manuscripts acquired in multispectral modality ", meeting title: 10th International Conference on Document Analysis and Recognition), this The major defect of method is that abatement effect is not ideal enough.

Summary of the invention

The present invention is a kind of method of page back side content cut down and appear in scanogram.The method includes following 6 Individual step.

1. white substrate and each run-down of black substrate on the page behind pad, obtain 2 width scanogram I_wAnd I_b。

2. detect obtain image I with image key points detection and description method_wAnd I_bOn key point and describe son, Then two groups of key points are mated with description.

3. by the key point matched to determining image I_wAnd I_bBetween the parameter of rigid body translation.

4. choose and labelling image I_wFront some pixels that upper brightness step modulus value is the highest.

5. set up image I_wAnd I_bBetween Linear Mapping f of pixel brightness value_I。

6. for image I_wOn each pixel, if its brightness value is less than image I_bThe f of the brightness value of upper corresponding point_I Inverse transformation value, then with the f of the pixel value of these corresponding point_IInverse transformation value substitutes；Otherwise keep constant.This Sample has just obtained the image after having cut down the back side content appeared.

Accompanying drawing explanation

Fig. 1, the use schematic diagram of the inventive method.Legend: 1 flatbed scanner；2 scanning windows；3 sweep Retouch the page；4 substrates.Illustrating: during scanning, the scanning page is close to the scanning window of scanner, and substrate is tight Against the page.

Scanogram after the upper white substrate of Fig. 2, page pad behind.

Scanogram after the upper black substrate of Fig. 3, page pad behind.

The gradient-norm visual image of the scanogram after the upper white substrate of Fig. 4, page pad behind, pixel intensity It is directly proportional to gradient-norm numerical value.

Fig. 5, cut down the scanogram after the back side content appeared.

Detailed description of the invention

Step 1:

Preparing white and each one piece of black substrate, size is slightly larger than the scanning window of scanner, it is desirable to surface is put down Whole, optical characteristics is uniform.The reflecting rate of white substrate the highest more good (more Bai Yuehao), black substrate reflective Rate the lowest more good (the most black more good).

White substrate and each run-down of black substrate (Fig. 1) on the page behind pad, obtain 2 width digitized maps As I_wAnd I_b, it is common that RGB (RGB) coloured image, i.e. each pixel is led to by 3 spectrum of RGB The luma component values in road is constituted, or single pass luminance picture (black white image).Fig. 2 and Fig. 3 shows Scanogram after the upper white substrate of example page pad behind and black substrate.

Step 2:

With the method for detection with the key point described on image, such as Scale Invariant Feature Transform (International Academic journal article, 2004, author: David G.Lowe, exercise question: " Distinctive Image Features from Scale-Invariant Keypoints ", journal title Claim: International Journal of Computer Vision), obtain image I_wAnd I_bOn key Point and description thereof, then mate two groups of key points according to the similarity described between son.

Step 3:

Use rigid body translation $\left(\begin{array}{c}{x}^{\′}\\ y^{\′}\end{array}\right)=\left(\begin{array}{cc}\cos \mathrm{\θ}& \sin \mathrm{\θ}\\ -\sin \mathrm{\θ}& \cos \mathrm{\θ}\end{array}\right)\left(\begin{array}{c}x\\ y\end{array}\right)+\left(\begin{array}{c}a\\ b\end{array}\right)$ Associated images I_wPlane and I_bPlane, its Parameter is (a, b, θ).

For often organizing 2 to match point, set up 1 about unknown number by these 2 pairs of coordinate vector A, b, sin θ, the linear equation system of cos θ

$\left(\begin{array}{cccc}{y}_1& x_1& 1& 0\\ -x_1& y_1& 0& 1\\ y_2& x_2& 1& 0\\ {-x}_2& y_2& 0& 1\end{array}\right)\left(\begin{array}{c}\sin \mathrm{\θ}\\ \cos \mathrm{\θ}\\ a\\ b\end{array}\right)=\left(\begin{array}{c}{x}_1^{\′}\\ y_1^{\′}\\ x_2^{\′}\\ y_2^{\′}\end{array}\right)$

Solve this system of linear equations, obtain the numerical value of these 4 unknown quantitys, then by (sin θ)²+ (cosθ)²=1 reasonability verifying solution, retains the solution vector (a, b, θ) by checking, then uses these Solve the average of vector as rigid body translation parameter.

Step 4:

Calculate image I_wThe brightness step vector of upper each pixelAccording to gradient vector mouldNumerical value, sort all pixels, labelling image I_wUpper brightness step modulus value is the highest Front some pixels.Fig. 4 shows the result of calculation of the gradient-norm of example.

Step 5:

According to the brightness value of pixel, ordering chart is as I_wOn, the high gradient mould pixel of removing step 4 labelling with Outer, other all pixels.

Choose the front P that brightness value is maximum_hiw% pixel, calculates the average brightness value of these pixelsUse step The rigid body translation parameter that 3 obtain is corresponding at image I to calculate these pixels_bOn coordinate, then with bilinearity insert Value method tries to achieve these images I_bOn the brightness value of corresponding point, then calculate its average brightness valueSimilarly, Choose the front P that numerical value is minimum_liw% pixel, calculates the average brightness value of these pixelsObtain by step 3 Rigid body translation parameter to calculate these pixels corresponding at image I_bOn coordinate, then ask with bilinear interpolation Obtain these images I_bOn the brightness value of corresponding point, then calculate its average brightness value

Then, with 2 to brightness number mapping relations, i.e.It is mapped toAndIt is mapped to Calculate brightness linear and map f_IThe parameter of ().

Step 6:

With image I_wBased on, for each pixel on image, if its brightness value is less than image I_bUpper correspondence The f of the brightness value that the bilinearity difference of point obtains_IInverse transformation value, then obtain by the bilinearity difference of these corresponding point The f of pixel value_IInverse transformation value substitutes；Otherwise keep constant.Thus obtain having cut down in the back side appeared Image after appearance.Scanogram after the back side content that the abatement of example page that shows Fig. 5 appears.

Claims (7)

1. the method cutting down the page back side content appeared in scanogram, it is characterised in that include following 6 steps Rapid: (one) is white substrate and each run-down of black substrate on the page behind pad, obtain 2 width scannings Image I_wAnd I_b.(2) detect obtain image I with image key points detection and description method_wAnd I_bOn Key point and description thereof, then mate two groups of key points with description.(3) with the pass matched Key point is to determining image I_wAnd I_bBetween the parameter of rigid body translation.(4) also labelling image I is chosen_wOn Front some pixels that the numerical value of brightness step mould is the highest.(5) image I is set up_wAnd I_bBetween pixel intensity Linear Mapping f of value_I(·).(6) for image I_wOn each pixel, if its brightness value is less than image I_bThe f of the brightness value of upper corresponding point_I() inverse transformation value, then with the f of the pixel value of these corresponding point_I() inverse transformation Value substitutes；Otherwise keep constant.

A kind of method cutting down the page back side content appeared in scanogram the most according to claim 1, it is special Levy and be white substrate and each run-down of black substrate on the page to be scanned behind pad, obtain 2 width and sweep Tracing is as I_wAnd I_b。

A kind of method cutting down the page back side content appeared in scanogram the most according to claim 1, it is special Levy and be to detect obtain image I with image key points detection and description method_wAnd I_bOn key point and Describe son, then mate two groups of key points with description.

A kind of method cutting down the page back side content appeared in scanogram the most according to claim 1, it is special Levy and be by the key point matched determining image I_wAnd I_bBetween the parameter of rigid body translation.

A kind of method cutting down the page back side content appeared in scanogram the most according to claim 1, it is special Levy and be to choose and labelling image I_wFront some pixels that the numerical value of upper brightness step mould is the highest.

A kind of method cutting down the page back side content appeared in scanogram the most according to claim 1, it is special Levy and be to set up image I_wAnd I_bBetween Linear Mapping f of pixel brightness value_I。

A kind of method cutting down the page back side content appeared in scanogram the most according to claim 1, it is special Levy and be for image I_wOn each pixel, if its brightness value is less than image I_bThe brightness of upper corresponding point The f of value_IInverse transformation value, then with the f of the pixel value of these corresponding point_IInverse transformation value substitutes；Otherwise keep not Become.