US 6,947,610 B2Grant

Method of correcting an image with perspective distortion and producing an artificial image with perspective distortion

Issue Date:2005-09-20

•5 Claims

•7 Drawing Sheets

Abstract

A method of correcting an image with perspective distortion and creating an artificial image with perspective distortion is disclosed. First, image data including a distorted object image are received. Next, the user inputs an operational instruction for setting a plurality of distortion parameters pertaining to the image data, which include a distorted side of the distorted object image, a central line and a distortion ratio. Thus, a plurality of coordinates of corners of the distorted object image are calculated according to these distortion parameters and a transformation matrix is determined according to the calculated coordinates of the corners of the distorted object image. Finally, the distorted object image is transformed to a corrected object image according to the transformation matrix.

Metadata

Assignee

Ulead Systems Inc.

Inventor

Hung-Ming Sun

Application Information

Application Number:US 10/085,409

Filing Date:2002-03-01

Priority Date:2001-03-02

Art Unit:2625

Patent Drawings (7 sheets)

Description

BACKGROUND OF THE INVENTION

[0001] 1. Field of the Invention

[0002] The present invention is related to an image processing technology, and more specifically, to a method of correcting an image with perspective distortion and producing an artificial image with perspective distortion.

[0003] 2. Description of the Prior Art

[0004] When an observer observes a physical object, the observed shape of the physical object is usually slightly different from the original one, which is determined by the distance between the observer and the object and the location of the observer. Such deviation is called perspective distortion. FIG. 1A (Prior Art) is a schematic diagram illustrating an observer and an object. FIG. 1B (Prior Art) is a schematic diagram illustrating an image with perspective distortion. As shown in FIG. 1A, the observer 8 observes the object 10 at a lower position, where numeral 6 represents a line of sight of the observer 8. In addition, the four corners of the object 10 are denoted by A1, A2, A3 and A4, respectively. Apparently, a line {overscore (A3A4)} is close to the observer 8 and a line {overscore (A1A2)} is distant from the observer 8. Accordingly, the observed lengths of the two lines {overscore (A3A4)} and {overscore (A1A2)} are extended and shortened, respectively, as shown in FIG. 1B. In FIG. 1B, the observed corners of the object 10 are denoted by A1′, A2′, A3′ and A4′.

[0005] In the image-processing field, the captured image of a physical object usually suffers from the distortion effect and should be corrected. A conventional correction scheme is described as follows. First, the user needs to locate the four corners of a distorted object on an image, which suffers from the perspective distortion effect. For example, as shown in FIG. 1B, the four corners A1′, A2′, A3′ and A4′ of the object 10 are determined by the user. Thus, the pixels of the image are transformed and the new coordinates of these pixels are calculated based on a coordinate-transforming matrix. Finally, the perspective distortion effect is corrected.

[0006] However, in the conventional scheme, the user must manually locate the corners of the distorted object and thus determine the corresponding transformation matrix, which is inconvenient for the user.

SUMMARY OF THE INVENTION

[0007] Therefore, the object of the present invention is to provide a method of processing an image with perspective distortion, which determines the corner coordinates with respect to various distorted sides of a distorted object.

[0008] The present invention achieves the above-indicated objects by providing a method of correcting an image with perspective distortion. First, image data including a distorted object image are received. Next, the user inputs an operational instruction for setting a plurality of distortion parameters pertaining to the image data, which include a distorted side of the distorted object image, a central line and a distortion ratio. Thus, a plurality of coordinates of corners of the distorted object image are calculated according to these distortion parameters and a transformation matrix is determined according to the calculated coordinates of the corners of the distorted object image. Finally, the distorted object image is transformed to a corrected object image according to the transformation matrix.

BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The following detailed description, given by way of example and not intended to limit the invention solely to the embodiments described herein, will best be understood in conjunction with the accompanying drawings, in which:

[0010] FIG. 1A (Prior Art) is a schematic diagram illustrating an observer and an object;

[0011] FIG. 1B (Prior Art) is a schematic diagram illustrating an image with perspective distortion;

[0012] FIG. 2 is a block diagram of processing an image with perspective distortion according to the preferred embodiment of the present invention;

[0013] FIG. 3 is a schematic diagram illustrating a first example of the correction of the perspective distortion when the distorted side is a top side according to the preferred embodiment of the present invention;

[0014] FIG. 4 is a schematic diagram illustrating a second example of the correction of the perspective distortion when the distorted side is a top side according to the preferred embodiment of the present invention;

[0015] FIG. 5 is a schematic diagram illustrating a first example of the correction of the perspective distortion when the distorted side is a top side according to the preferred embodiment of the present invention;

[0016] FIG. 6 is a schematic diagram illustrating the determination of the corner coordinates of the distorted object image with a distorted bottom side according to the preferred embodiment of the present invention; and

[0017] FIG. 7 is a schematic diagram illustrating the determination of the corner coordinates of the distorted object image with a distorted left side according to the preferred embodiment of the present invention.

DETAILED DESCRIPTION OF THE INVENTION

[0018] FIG. 2 is a block diagram of processing an image with perspective distortion in the embodiment. Image data 20 shown in FIG. 2 include a distorted object image ready to be processed, such as the distorted object shown in FIG. 1B. Data image 20 are sent to a preprocessing unit 30, which is used to receive the distortion parameters that are required in the correction of perspective distortion. There are three distortion parameters in the preferred embodiment, including a distortion side, a location of a distortion central line and a distortion ratio. The distorted side is denoted by “D” and indicates the distorted side of the distorted object image that is ready to be processed. The distorted side D can be set as TOP, BOTTOM, LEFT and RIGHT, which indicate the top side, the bottom side, the left side and right side of the distorted object image, respectively. The location of the distortion central line is denoted by “C” and will be described later in details. The distortion ratio is denoted by “F” and indicates a ratio of the length of the distorted side and the length of the original side with respect to the distorted side. The distortion ratio is between 0 and 1 (no distortion).

[0019] When these distortion parameters are prepared, the image data 20 is sent to a coordinate-determining unit 40 for determining the coordinates of the corners of the distorted object image and a corresponding transformation matrix. Finally, a transforming unit 50 transforms the distorted object image to an undistorted image according to the transformation matrix. It is noted that an artificial image with perspective distortion can also be produced by a similar process. In the preferred embodiment, all distorted sides of the distorted object image can be sequentially processed by setting different sets of the distortion parameters to determine the required transformation matrix.

[0020] The following description focuses on the determination of the coordinates of the corners of the distorted object image. As described above, the distortion side D can be set as TOP, BOTTOM, LEFT and RIGHT. The four cases will be described respectively in details.

The Distortion Side D is TOP (D=TOP)

First Example

[0021] FIG. 3 is a schematic diagram illustrating a first example when the distorted side is TOP according to the preferred embodiment of the present invention. The four corners of the distorted object image are denoted by s, C_11t, C_12tand r, respectively. The coordinates of the two corners C_11tand C_12tare undetermined since the distortion side D is a top side. In addition, the two corners s and r and another two points p and q constitute four corners of a rectangle, which is illustrated as dash lines. The coordinates of the four points p, q, r and s are defined as p=(0,0), q=(W,0), r=(W,H) and s=(0,H), where the values W and H are positive numbers. The central line C_1tof the present embodiment is determined as follows. The intersection of extensions of the two sides {overscore (sC_11t)} and {overscore (rC_12t)} is defined as a point Z₁. Thus, the central line C_1tis a line perpendicular to the side {overscore (C_11tC_12t)} passing through the point Z₁, where the intersection of the central line C_1tand the side {overscore (C_11tC_12t)} is E₁. The first example shown in FIG. 3 illustrates the situation when the central line C_1tpasses through the inside of the distorted object image.

[0022] The coordinates of the corners s, C_11t, C_12tand r are determined by the known distorted side (D=TOP), the central line C_1tand the distortion ratio F. As described above, the coordinates of the corners s and r are (0,H) and (W,H), respectively. The coordinates of the corners C_11tand C_12tare determined as follows.

[0023] With respect to the corner C_11t, its y coordinate is the same as the y coordinates of the points p and q, and its x coordinate is calculated by:
C_11t,x=x_1t−x_1t×F (1)

[0024] where x_1trepresents the x coordinate of the central line C_1t. The complete coordinates of the corner C_11tare obtained.

[0025] With respect to the corner C_12t, its y coordinate is the same as the y coordinates of the points p and q, and its x coordinate is acquired by using the relationship of the similar triangles shown in FIG. 3.

\begin{matrix} \frac{\overline{{pC}_{11 t}}}{\overline{C_{11 t} E_{1}}} = \frac{\overline{C_{11 t} s}}{\overline{C_{11 t} Z_{1}}} = \frac{\overline{C_{12 t} r}}{\overline{C_{12 t} Z_{1}}} = \frac{\overline{{qC}_{12 t}}}{\overline{C_{12 t} E_{1}}} & (2) \end{matrix}

[0026] According to equations (1) and (2), the x coordinate of the corner C_12tis calculated by:

\begin{matrix} \begin{matrix} C_{12 t, x} = x_{it} + \overline{{qE}_{1}} \times F \\ = x_{1 t} + (W - x_{1 t}) \times F \end{matrix} & (3) \end{matrix}

[0027] where x_1trepresents the x coordinate of the central line C_1t. The complete coordinates of the corner C_12tare obtained.

[0028] Therefore, the complete coordinates of the four corners r, C_11t, C_12tand s of the distorted object image are obtained in the first example.

Second Example

[0029] FIG. 4 is a schematic diagram illustrating a second example when the distorted side is TOP according to the preferred embodiment of the present invention. The four corners of the distorted object image are denoted by p, C_22t, r and C_21t, respectively. The coordinates of the two corners C_22tand C_21tare undetermined. In addition, the two corners p and r and another two points q and s constitute four corners of a rectangle, which is illustrated as dash lines. The coordinates of these corners of the rectangular can be defined as p=(0,0), q=(W,0), r=(W,H) and s=(0,H), which are the same as those in the first example. In the second example, the central line C_2tis located at the left of the distorted object image and determined as follows. The intersection of extensions of the two sides {overscore (C_21tp)} and {overscore (rC_22t)} is defined as a point Z₂. Thus, the central line C_2tis a line perpendicular to an extension of the side {overscore (C_22tp)} passing through the point Z₂, where the intersection of the central line C_2tand the extension of the side {overscore (C_22tp)} is E₂. In addition, the length of the line segment {overscore (E₂Z₂)} is denoted by l_2t.

[0030] The coordinates of the corners p, C_22t, r and C_21tare determined by the known distorted side (D=TOP), the central line C_2tand the distortion ratio F. As described above, the coordinates of the corners p and r are (0,0) and (W,H), respectively. The coordinates of the corners C_21tand C_22tare determined as follows.

[0031] With respect to the corner C_22t, its y coordinate is the same as the y coordinate of the point p (that is, 0), and its x coordinate is calculated by:
C_22t,x=x_2t+(W+x_2t)×F (4)

[0032] where x_2tis a negative number and represents the x coordinate of the central line C_2t. The complete coordinates of the corner C_22tare obtained.

[0033] With respect to the corner C_21t, its y coordinate is the same as the y coordinate of the point s (that is, H), and its x coordinate is calculated as follows. The angle of an extension of the side {overscore (rC_22t)} and the central line C_2tis defined as θ_2t. In addition, the angle of the line segments {overscore (rq)} and {overscore (rC_22t)} is also θ_2t. According to FIG. 4,

\tan θ_{2 t} = \frac{\overline{C_{22 t} q}}{\overline{rq}} = \frac{W - C_{22 t, x}}{H}

[0034] Thus, the angle θ_2tis obtained by the following equation:

\begin{matrix} θ_{2 t} = \tan^{- 1} (\frac{W - C_{22 t, x}}{H}) & (5) \end{matrix}

[0035] In addition, the length l_2tis obtained by the following equation:

\begin{matrix} \begin{matrix} ℓ_{2 t} = \overline{C_{22 t} E_{2}} \cot θ_{2 t} \\ = [\langle x_{2 t} \rangle + W - (W - C_{22 t, x})] \cot θ_{2 t} \\ = (C_{22 t, x} - x_{2 t}) \cot θ_{2 t} \end{matrix} & (6) \end{matrix}

[0036] where x_2trepresents the x coordinate of the central line C_2t. In addition, since triangles Z₂E₂p and psC_21tare similar, the following equation is satisfied.

\begin{matrix} \frac{ℓ_{2 t}}{\langle x_{2 t} \rangle} = \frac{\overline{p s}}{C_{21 t, x}} & (7) \end{matrix}

[0037] The x coordinate of the corner C_21tcan be determined by equations (5), (6) and (7).
C_21t,x=|x_2t|×H/[(C_22t,x−x_2t)cot θ_2t] (8)

[0038] Accordingly, the complete coordinates of the corner C_22tare obtained by equation (4) and the complete coordinates of the corner C_21tare obtained by equation (8).

Third Example

[0039] FIG. 5 is a schematic diagram illustrating a third example when the distorted side is TOP according to the preferred embodiment of the present invention. The four corners of the distorted object image are denoted by C_31t, q, C_32tand s, respectively. The coordinates of the two corners C_31tand C_32tare undetermined. In addition, the two corners q and s and another two points p and r constitute a rectangle, which is illustrated as dash lines. The coordinates of these corners of the rectangular can be defined as p=(0,0), q=(W, 0), r=(W,H) and s=(0,H), which are the same as those in the first example. In the third example, the central line C_3tis located at the right of the distorted object image and determined as follows. The intersection of extensions of the two sides {overscore (sC_31t)} and {overscore (C_32tp)} is defined as a point Z₃. Thus, the central line C_3tis a line perpendicular to an extension of the side {overscore (C_31tq)} passing through the point Z₃, where the intersection of the central line C_2tand the extension of the side {overscore (C_31tq)} is E₃. In addition, the length of the line segment {overscore (E₃Z₃)} is denoted by l_3t.

[0040] The coordinates of the corners C_31t, q, C_32tand s are determined by the known distorted side (D=TOP), the central line C_3tand the distortion ratio F. As described above, the coordinates of the corners q and s are (W, 0) and (0,H), respectively. The coordinates of the corners C_31tand C_32tare determined as follows.

[0041] With respect to the corner C_31t, its y coordinate is the same as the y coordinate of the point q (that is, 0), and its x coordinate is calculated by:
C_31t,x=x_3t−x_3t×F (9)

[0042] where x_3trepresents the x coordinate of the central line C_3t. The complete coordinates of the corner C_31tare obtained.

[0043] With respect to the corner C_32t, its y coordinate is the same as the y coordinate of the point s (that is, H), and its x coordinate is calculated as follows. The angle of an extension of the side {overscore (sC_31t)} and the central line C_3tis defined as θ_3t. In addition, the angle of the line segments {overscore (sp)} and {overscore (pC_31t)} is also θ_3t. As shown in FIG. 5,

\tan θ_{3 t} = \frac{\overline{{pC}_{31 t}}}{\overline{sp}} = \frac{C_{31 t, x}}{H}

[0044] Thus, the angle θ_3tis obtained by the following equation:

\begin{matrix} θ_{3 t} = \tan^{- 1} (\frac{C_{31 t, x}}{H}) & (10) \end{matrix}

[0045] In addition, the length l_3tis obtained by the following equation:

\begin{matrix} \begin{matrix} ℓ_{3 t} = \overline{E_{3} C_{31 t}} \cot θ_{3 t} \\ = (x_{3 t} - C_{31 t, x}) \cot θ_{3 t} \\ = x_{3 t} \times F \times \cot θ_{3 t} \end{matrix} & (11) \end{matrix}

[0046] Since triangles Z₃E₃q and qrC_32tare similar, the following equation is satisfied.

\begin{matrix} \frac{\overline{{rC}_{32 t}}}{\overline{qr}} = \frac{E_{3} q}{ℓ_{3 t}} & (12) \end{matrix}

[0047] The x coordinate of the corner C_32tcan be determined by equations (10), (11) and (12).

\begin{matrix} \begin{matrix} C_{32 t, x} = W - \overline{{rC}_{32 t}} \\ = W - \frac{H \times (x_{3 t} - W)}{x_{3 t} \times F \times \cot θ_{3 t}} \end{matrix} & (13) \end{matrix}

[0048] Accordingly, the complete coordinates of the corner C_31tare obtained by equation (9) and the complete coordinates of the corner C_32tare obtained by equation (13).

[0049] According to the three examples described above, the coordinates of the corners of the distorted object image can be determined by the known distortion parameters when the distortion side is a top side. Subsequently, a new transformation matrix can be obtained and the distorted object image is transformed by means of the transformation matrix.

The Distortion Side D is BOTTOM (D=BOTTOM)

[0050] FIG. 6 is a schematic diagram illustrating the determination of the corner coordinates of the distorted object image with a distorted bottom side according to the preferred embodiment of the present invention. In addition, FIG. 6 shows three different central lines C_1b(passing through the inside of the distorted object image), C_2b(locating at the left of the distorted object image) and C_3b(locating at the right of the distorted object image), which correspond to three examples in this case.

[0051] In the first example, the central line is denoted by C_1b. The four corners of the distorted object image are denoted by p, q, C_11band C_12b, respectively. The coordinates of the two corners C_11band C_12bare undetermined. In addition, the other corners p and q and two points r and s constitute four corners of a rectangle, which is illustrated as dash lines, where p=(0,0), q=(W,0), r=(W,H) and s=(0,H). The central line C_1bis determined by the intersection Z₄of extensions of the two sides {overscore (pC_11b)} and {overscore (qC_12b)}. The central line C_1bis a line perpendicular to the side {overscore (C_11bC_12b)} passing through the point Z₄, where the intersection of the central line C_1band the side {overscore (C_11bC_12b)} is E₄.

[0052] The coordinates of the corners p, q, C_11band C_12bare determined by the known distorted side (D=BOTTOM), the central line C_1band the distortion ratio F. As described above, the coordinates of the corners p and g are (0,0) and (W,0), respectively. The coordinates of the corners C_11band C_12bare determined as follows.

[0053] With respect to the corner C_11b, its y coordinate is the same as the y coordinates of the points s and r, and its x coordinate is calculated by:
C_11b,x=x_1b−x_1b×F (14)

[0054] where x_1brepresents the x coordinate of the central line C_1b. The complete coordinates of the corner C_11bare obtained.

[0055] With respect to the corner C_12b, its y coordinate is the same as the y coordinates of the points s and r. According to FIG. 6, the following equation is satisfied:

\begin{matrix} \frac{\overline{{sC}_{11 b}}}{\overline{C_{11 b} E_{4}}} = \frac{\overline{C_{11 b} p}}{\overline{C_{11 b} Z_{4}}} = \frac{\overline{C_{12 b} q}}{\overline{C_{12 b} Z_{4}}} = \frac{\overline{{rC}_{12 b}}}{\overline{C_{12 b} E_{4}}} & (15) \end{matrix}

[0056] According to equations (14) and (15), the x coordinate of the corner C_12bis calculated by:

\begin{matrix} \begin{matrix} C_{12 b, x} = x_{1 b} + \overline{{rE}_{4}} \times F \\ = x_{1 b} + (W - x_{1 b}) \times F \end{matrix} & (16) \end{matrix}

[0057] where x_1brepresents the x coordinate of the central line C_1b. The complete coordinates of the corner C_12bare obtained by equation (16).

[0058] The determination of the corner coordinates in the second example, in which the central line is located at the left of the distorted object image, and the third example, in which the central line is located at the right of the distortion object image are similar to those in the first embodiment.

[0059] Accordingly, the coordinates of the corners of the distorted object image can be determined by the known distortion parameters when the distortion side is a BOTTOM side. Subsequently, a new transformation matrix can be obtained and the distorted object image is transformed by means of the transformation matrix.

The Distortion Side D is LEFT (D=LEFT)

[0060] FIG. 7 is a schematic diagram illustrating the determination of the corner coordinates of the distorted object image with a distorted left side according to the preferred embodiment of the present invention. FIG. 7 shows three different central lines C_1l(passing through the inside of the distorted object image), C_2l(locating above the distorted object image) and C_3l(locating below the distorted object image), which correspond to three examples in this situation.

First Example

[0061] The central line in the first example is denoted by C_1l. The four corners of the distorted object image are denoted by q, r, C_12land C_11l, respectively, where the coordinates of C_12land C_11lare undetermined. In addition, the two corners s and r and another two points p and q constitute four corners of a rectangle, where p=(0,0), q=(W,0), r=(W,H) and s=(0,H). The central line C_1lis determined by the intersection Z₇of extensions of the two sides {overscore (qC_11l)} and {overscore (rC_12l)}. Thus, the central line C_1lis a line perpendicular to the side {overscore (C_11lC_12l)} passing through the point Z₇, where the intersection of the central line C_1land the side {overscore (C_11lC_12l)} is E₇.

[0062] The coordinates of the corners q, r, C_12land C_11lare determined by the known distorted side (D=LEFT), the central line C_1land the distortion ratio F. As described above, the coordinates of the corners q and r are (W,0) and (W,H), respectively. The coordinates of the corners C_11land C_12lare determined as follows.

[0063] With respect to the corner C_11l, its x coordinate is the same as the x coordinates of the points p and s, and its y coordinate is calculated by:
C_11l,y=y_1l−y_1l×F (17)

[0064] where y_1lrepresents the y coordinate of the central line C_1l. The complete coordinates of the corner C_11lare obtained.

[0065] With respect to the corner C_12l, its x coordinate is the same as the x coordinates of the points p and s, and its x coordinate is acquired by using the relationship of the similar triangles, which is similar to that in the case of D=TOP.

\begin{matrix} \begin{matrix} C_{12 l, y} = y_{1 l} + \overline{{sE}_{7}} \times F \\ = y_{1 l} + (H - y_{1 l}) \times F \end{matrix} & (18) \end{matrix}

[0066] where y_1lrepresents the y coordinate of the central line C_1l. The complete coordinates of the corner C_12lare obtained.

[0067] Therefore, the complete coordinates of the corner C_11lare obtained by equation (17) and the complete coordinates of the corner C_12lare obtained by equation (18).

Second Example

[0068] The central line in the first example is denoted by C_2land located above the distorted object image. The four corners of the distorted object image are denoted by p, C_21l, r and C_22l, respectively, where the coordinates of the two corners C_21land C_22lare undetermined. The central line C_2lis determined by the intersection Z₈of extensions of the two sides {overscore (C_21lp)} and {overscore (rC_22l)}. Thus, the central line C_2lis a line perpendicular to an extension of the side {overscore (C_22lp)} passing through the point Z₈, where the intersection of the central line C_2land the extension of the side {overscore (C_22lp)} is E₈. In addition, the length of the line segment {overscore (E₈Z₈)} is denoted by l_2l.

[0069] With respect to the corner C_22l, its x coordinate is the same as the x coordinate of the point p (that is, 0) and its y coordinate is calculated by:
C_22l,y=y_2l+(H−y_2l)×F (19)

[0070] where y_2lis a negative number and represents the y coordinate of the central line C_2l.

[0071] With respect to the corner C_21l, its x coordinate is the same as the x coordinate of the point r (that is, W), and its y coordinate is calculated by the same method shown in FIG. 4.
C_21l,y=|y_2l|×W/[(C_22l,y−y_2l)cot θ_2l] (20)

[0072] where y_2lrepresents the y coordinate of the central line C_2land θ_2lis an angle of the line segments {overscore (rC_22l)} and {overscore (rs)}. Accordingly, the complete coordinates of the corner C_22lare obtained by equation (19) and the complete coordinates of the corner C_21lare obtained by equation (20).

Third Example

[0073] The central line in the first example is denoted by C_3land located below the distorted object image. The four corners of the distorted object image are denoted by q, C_32l, s and C_31lrespectively, where the coordinates of the two corners C_32land C_31lare undetermined. The central line C_3lis determined by the intersection Z₉of extensions of the two sides {overscore (C_32ls)} and {overscore (qC_31l)}. Thus, the central line C_3lis a line perpendicular to an extension of the side {overscore (C_31ls)} passing through the point Z₉, where the intersection of the central line C_3land the extension of the side {overscore (C_31ls)} is E₉. In addition, the length of the line segment {overscore (E₉Z₉)} is denoted by l_3l.

[0074] With respect to the corner C_31l, its x coordinate is the same as the x coordinate of the point s (that is, 0), and its y coordinate is calculated by:
C_31l,y=y_3l−y_3l×F (21)

[0075] where y_3lrepresents the y coordinate of the central line C_3l.

[0076] With respect to the corner C_32l, its x coordinate is the same as the x coordinate of the point q (that is, W), and its y coordinate is calculated by the same method shown in FIG. 5.

\begin{matrix} \begin{matrix} C_{32 ℓ, y} = H - \overline{{rC}_{32 ℓ}} \\ = H - \frac{W \times (y_{3 ℓ} - H)}{y_{3 ℓ} \times F \times \cot θ_{3 ℓ}} \end{matrix} & (22) \end{matrix}

[0077] where y_3lrepresents the y coordinate of the central line C_3land θ_3lis an angle of the line segments {overscore (qC_31l)} and {overscore (qp)}. Accordingly, the complete coordinates of the corner C_31lare obtained by equation (21) and the complete coordinates of the corner C_32lare obtained by equation (22)

[0078] According to the three examples described above, the coordinates of the corners of the distorted object image can be determined by the known distortion parameters when the distortion side is the left side of the distorted object image. Subsequently, a new transformation matrix can be obtained and the distorted object image is transformed by means of the transformation matrix.

The Distortion Side D is LEFT (D=LEFT)

[0079] The process of the determination of the corner coordinates of the distorted object image in this case, in which the distortion side is a left side of the distorted object image, is similar to those in the above cases and will not be described again.

[0080] It is noted that the above-described process also can be applied to produce an artificial image with perspective distortion, where the original object image is undistorted.

[0081] While the invention has been described by way of example and in terms of the preferred embodiment, it is to be understood that the invention is not limited to the disclosed embodiments. On the contrary, it is intended to cover various modifications and similar arrangements as would be apparent to those skilled in the art. Therefore, the scope of the appended claims should be accorded the broadest interpretation so as to encompass all such modifications and similar arrangements.

Claims

1. A method of correcting an image with perspective distortion, comprising the steps of:

receiving image data including a distorted object image, wherein the distorted object image is perspective from a projection plane;

setting distortion parameters pertaining to the image data according to an operational instruction, wherein the distortion parameters comprise a distortion central line perpendicular to the projection plane and passing through a coverage point of any set of parallel lines that are not parallel to the projection plane;

calculating a plurality of coordinates of corners of the distorted object image according to the distortion parameters;

determining a transformation matrix according to the calculated coordinates of the corners of the distorted object image; and

transforming the distorted object image to a corrected object image according to the transformation matrix.

2. The method of claim 1, wherein the distortion parameter includes a distortion ratio of a distorted side of the distorted object image.

3. The method of claim 1, wherein the coordinates of the corners of the corrected object image are calculated according to the location of the distortion central line of the distorted object image.

4. The method of claim 1, wherein the distortion parameter includes a distorted side of the distorted object image, a distortion ratio of the distorted side of the distorted object image and a location of a distortion central line of the distorted object image.

5. A method of determining coordinates of corners of a distorted object image on an image with perspective distortion, comprising the steps of:

determining a distorted side of the distorted object image, wherein the distorted object image is perspective from a projection plane;

determining a location of a distortion central line of the distorted object image with respect to the distorted side and a distortion ratio of the distorted object image with respect to the distorted side, wherein the distortion central line is perpendicular to the projection plane and passes through a coverage point of any set of parallel lines that are not parallel to the projection plane; and

calculating the coordinates of the corners of the distorted object image according to the distorted side, the location of the distortion central line and the distortion ratio.

Patent Citations (9)

Patent	Date	Inventor	Cited By
US5528194(A)	1996-06-01	Ohtani et al.	Examiner
US5726716(A)	1998-03-01	Egashira et al.	Examiner
US5933132(A)	1999-08-01	Marshall et al.	Examiner
US6324297(B1)	2001-11-01	Uchida	Examiner
US6361171(B1)	2002-03-01	Ejiri et al.	Examiner
US6450647(B1)	2002-09-01	Takeuchi	Examiner
US6511185(B1)	2003-01-01	Gananathan	Examiner
US6520647(B2)	2003-02-01	Raskar	Examiner
US6791616(B2)	2004-09-01	Tamaki et al.	Examiner

Non-Patent Literature (1)

Kannappan, IEEE Publication, 1992, “ An interleaved scanline algorithm for 2-D affine transformations of images”. (pp. 179-182).Examiner

Abstract

Description

BACKGROUND OF THE INVENTION

[0001] 1. Field of the Invention

[0003] 2. Description of the Prior Art

SUMMARY OF THE INVENTION

BRIEF DESCRIPTION OF THE DRAWINGS

[0010] FIG. 1A (Prior Art) is a schematic diagram illustrating an observer and an object;

[0011] FIG. 1B (Prior Art) is a schematic diagram illustrating an image with perspective distortion;

[0012] FIG. 2 is a block diagram of processing an image with perspective distortion according to the preferred embodiment of the present invention;

DETAILED DESCRIPTION OF THE INVENTION

The Distortion Side D is TOP (D=TOP)

First Example

[0023] With respect to the corner C_11t, its y coordinate is the same as the y coordinates of the points p and q, and its x coordinate is calculated by:
C_11t,x=x_1t−x_1t×F (1)

[0024] where x_1trepresents the x coordinate of the central line C_1t. The complete coordinates of the corner C_11tare obtained.

\begin{matrix} \frac{\overline{{pC}_{11 t}}}{\overline{C_{11 t} E_{1}}} = \frac{\overline{C_{11 t} s}}{\overline{C_{11 t} Z_{1}}} = \frac{\overline{C_{12 t} r}}{\overline{C_{12 t} Z_{1}}} = \frac{\overline{{qC}_{12 t}}}{\overline{C_{12 t} E_{1}}} & (2) \end{matrix}

[0026] According to equations (1) and (2), the x coordinate of the corner C_12tis calculated by:

\begin{matrix} \begin{matrix} C_{12 t, x} = x_{it} + \overline{{qE}_{1}} \times F \\ = x_{1 t} + (W - x_{1 t}) \times F \end{matrix} & (3) \end{matrix}

[0027] where x_1trepresents the x coordinate of the central line C_1t. The complete coordinates of the corner C_12tare obtained.

[0028] Therefore, the complete coordinates of the four corners r, C_11t, C_12tand s of the distorted object image are obtained in the first example.

Second Example

[0031] With respect to the corner C_22t, its y coordinate is the same as the y coordinate of the point p (that is, 0), and its x coordinate is calculated by:
C_22t,x=x_2t+(W+x_2t)×F (4)

[0032] where x_2tis a negative number and represents the x coordinate of the central line C_2t. The complete coordinates of the corner C_22tare obtained.

\tan θ_{2 t} = \frac{\overline{C_{22 t} q}}{\overline{rq}} = \frac{W - C_{22 t, x}}{H}

[0034] Thus, the angle θ_2tis obtained by the following equation:

\begin{matrix} θ_{2 t} = \tan^{- 1} (\frac{W - C_{22 t, x}}{H}) & (5) \end{matrix}

[0035] In addition, the length l_2tis obtained by the following equation:

\begin{matrix} \begin{matrix} ℓ_{2 t} = \overline{C_{22 t} E_{2}} \cot θ_{2 t} \\ = [\langle x_{2 t} \rangle + W - (W - C_{22 t, x})] \cot θ_{2 t} \\ = (C_{22 t, x} - x_{2 t}) \cot θ_{2 t} \end{matrix} & (6) \end{matrix}

[0036] where x_2trepresents the x coordinate of the central line C_2t. In addition, since triangles Z₂E₂p and psC_21tare similar, the following equation is satisfied.

\begin{matrix} \frac{ℓ_{2 t}}{\langle x_{2 t} \rangle} = \frac{\overline{p s}}{C_{21 t, x}} & (7) \end{matrix}

[0037] The x coordinate of the corner C_21tcan be determined by equations (5), (6) and (7).
C_21t,x=|x_2t|×H/[(C_22t,x−x_2t)cot θ_2t] (8)

[0038] Accordingly, the complete coordinates of the corner C_22tare obtained by equation (4) and the complete coordinates of the corner C_21tare obtained by equation (8).

Third Example

[0041] With respect to the corner C_31t, its y coordinate is the same as the y coordinate of the point q (that is, 0), and its x coordinate is calculated by:
C_31t,x=x_3t−x_3t×F (9)

[0042] where x_3trepresents the x coordinate of the central line C_3t. The complete coordinates of the corner C_31tare obtained.

\tan θ_{3 t} = \frac{\overline{{pC}_{31 t}}}{\overline{sp}} = \frac{C_{31 t, x}}{H}

[0044] Thus, the angle θ_3tis obtained by the following equation:

\begin{matrix} θ_{3 t} = \tan^{- 1} (\frac{C_{31 t, x}}{H}) & (10) \end{matrix}

[0045] In addition, the length l_3tis obtained by the following equation:

\begin{matrix} \begin{matrix} ℓ_{3 t} = \overline{E_{3} C_{31 t}} \cot θ_{3 t} \\ = (x_{3 t} - C_{31 t, x}) \cot θ_{3 t} \\ = x_{3 t} \times F \times \cot θ_{3 t} \end{matrix} & (11) \end{matrix}

[0046] Since triangles Z₃E₃q and qrC_32tare similar, the following equation is satisfied.

\begin{matrix} \frac{\overline{{rC}_{32 t}}}{\overline{qr}} = \frac{E_{3} q}{ℓ_{3 t}} & (12) \end{matrix}

[0047] The x coordinate of the corner C_32tcan be determined by equations (10), (11) and (12).

\begin{matrix} \begin{matrix} C_{32 t, x} = W - \overline{{rC}_{32 t}} \\ = W - \frac{H \times (x_{3 t} - W)}{x_{3 t} \times F \times \cot θ_{3 t}} \end{matrix} & (13) \end{matrix}

[0048] Accordingly, the complete coordinates of the corner C_31tare obtained by equation (9) and the complete coordinates of the corner C_32tare obtained by equation (13).

The Distortion Side D is BOTTOM (D=BOTTOM)

[0053] With respect to the corner C_11b, its y coordinate is the same as the y coordinates of the points s and r, and its x coordinate is calculated by:
C_11b,x=x_1b−x_1b×F (14)

[0054] where x_1brepresents the x coordinate of the central line C_1b. The complete coordinates of the corner C_11bare obtained.

[0055] With respect to the corner C_12b, its y coordinate is the same as the y coordinates of the points s and r. According to FIG. 6, the following equation is satisfied:

\begin{matrix} \frac{\overline{{sC}_{11 b}}}{\overline{C_{11 b} E_{4}}} = \frac{\overline{C_{11 b} p}}{\overline{C_{11 b} Z_{4}}} = \frac{\overline{C_{12 b} q}}{\overline{C_{12 b} Z_{4}}} = \frac{\overline{{rC}_{12 b}}}{\overline{C_{12 b} E_{4}}} & (15) \end{matrix}

[0056] According to equations (14) and (15), the x coordinate of the corner C_12bis calculated by:

\begin{matrix} \begin{matrix} C_{12 b, x} = x_{1 b} + \overline{{rE}_{4}} \times F \\ = x_{1 b} + (W - x_{1 b}) \times F \end{matrix} & (16) \end{matrix}

[0057] where x_1brepresents the x coordinate of the central line C_1b. The complete coordinates of the corner C_12bare obtained by equation (16).

The Distortion Side D is LEFT (D=LEFT)

First Example

[0063] With respect to the corner C_11l, its x coordinate is the same as the x coordinates of the points p and s, and its y coordinate is calculated by:
C_11l,y=y_1l−y_1l×F (17)

[0064] where y_1lrepresents the y coordinate of the central line C_1l. The complete coordinates of the corner C_11lare obtained.

\begin{matrix} \begin{matrix} C_{12 l, y} = y_{1 l} + \overline{{sE}_{7}} \times F \\ = y_{1 l} + (H - y_{1 l}) \times F \end{matrix} & (18) \end{matrix}

[0066] where y_1lrepresents the y coordinate of the central line C_1l. The complete coordinates of the corner C_12lare obtained.

[0067] Therefore, the complete coordinates of the corner C_11lare obtained by equation (17) and the complete coordinates of the corner C_12lare obtained by equation (18).

Second Example

[0069] With respect to the corner C_22l, its x coordinate is the same as the x coordinate of the point p (that is, 0) and its y coordinate is calculated by:
C_22l,y=y_2l+(H−y_2l)×F (19)

[0070] where y_2lis a negative number and represents the y coordinate of the central line C_2l.

Third Example

[0074] With respect to the corner C_31l, its x coordinate is the same as the x coordinate of the point s (that is, 0), and its y coordinate is calculated by:
C_31l,y=y_3l−y_3l×F (21)

[0075] where y_3lrepresents the y coordinate of the central line C_3l.

[0076] With respect to the corner C_32l, its x coordinate is the same as the x coordinate of the point q (that is, W), and its y coordinate is calculated by the same method shown in FIG. 5.

\begin{matrix} \begin{matrix} C_{32 ℓ, y} = H - \overline{{rC}_{32 ℓ}} \\ = H - \frac{W \times (y_{3 ℓ} - H)}{y_{3 ℓ} \times F \times \cot θ_{3 ℓ}} \end{matrix} & (22) \end{matrix}