KR20080053172A - Apparatus and method for detecting horizon in a sea image - Google Patents

Abstract

An apparatus and a method for extracting a horizontal line are provided to obtain quite stable horizontal extraction results, and easily extract horizontal lines from ocean scene image sequences of hundreds of frames due to automating horizontal line extraction. An interested region extracting unit(18) extracts coordinates of an interested region selected from a still image corresponding to each frame of ocean scene video. A controller(20) acquires a horizontal line corresponding to line adjustment from the coordinates of the interested region of the still image. The still image is adjusted in its contrast and Gaussian-filtered. The contrast is adjusted based on previous settings, and adjusted under the control of the controller. The interested region extracting unit extracts the coordinates of the interested region of the first still image among still images corresponding to each frame.

Description

Horizon extraction device and method {APPARATUS AND METHOD FOR DETECTING HORIZON IN A SEA IMAGE}

The present invention relates to an apparatus and method for extracting horizontal lines, and more particularly, to an apparatus and method for extracting horizontal lines necessary for obtaining camera movement in a marine scene image synthesis in a marker-free marine camera tracking system. It is about.

The present invention is derived from the research conducted as part of the IT new growth engine core technology development project of the Ministry of Information and Communication and the Institute of Information and Telecommunication Research and Development. Authoring software].

The synthesis of computer graphics (CG) and photorealism has made strides in the film industry and augmented reality. For natural compositing, we need to know the camera's movement when the video was shot. In general, humans manually compositing every frame, but this is very laborious and time consuming.

Many commercial software (shakes, carelessness, match movers, combuses, etc.) is used instead of manual work in image synthesis. After extracting and matching feature points of adjacent images in a sequence of images, the relative movements of the cameras are calculated through their correlations. However, in the video where feature points such as maritime scenes cannot be obtained, buoys were placed on the sea as in the movie “Troy,” and camera movements were obtained using them as feature points. In this case, the video compositing work is performed in most cases, such as the buoys that need to be connected to other ships during filming and floating on the sea, and when the buoys are moved by waves when close-ups are taken from other ships. Hard.

In a marine scene, information about the camera's tilt and yaw can be obtained using horizontal information without any markers. In a typical maritime scene, a horizontal line can be extracted using radon or hough transform that extracts a straight line, but when the weather conditions are bad, the horizontal line appears faint due to water fog. It is difficult to extract.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problem, and specifies a region of interest (ROI) near a horizontal line in a provided marine scene video, and corresponds to a pixel corresponding to one row in the region of interest. The purpose of the present invention is to provide a horizontal line extracting device and a method for extracting horizontal lines through line fitting using the maximum points found by performing the process of finding the largest brightness difference by using the brightness difference between the rows. .

In order to achieve this purpose,

According to one embodiment of the present invention, an apparatus for extracting a horizontal line from a maritime scene, comprising: a region of interest extracting unit for selecting a region of interest in a still image corresponding to every frame of a maritime scene video; And a controller for obtaining a horizontal line corresponding to a line alignment in the ROI among the still images.

Preferably, the still image may be contrast-adjusted and Gaussian filtered. The contrast is preferably adjusted based on setting. The contrast may be adjusted according to the control of the controller.

The region of interest extractor may select the region of interest from the first still image of the still image corresponding to each frame.

The line alignment calculates a difference between contrast values of two pixels that are vertically symmetric about each pixel in each column in the ROI, and selects maximum points corresponding to pixels having the largest contrast difference in each column. Calculations are preferably performed using the maximum points. The column is preferably all columns in the region of interest.

The control unit may assign the same file name to a plurality of still images corresponding to the sea scene video, and sequentially assign a serial number to each still image so that each still image may be collectively processed as one file. .

The ROI selection of the controller may be performed by receiving the upper left point and the lower right point from the ROI extractor.

According to another aspect of the present invention, there is provided a horizontal line extracting method in an apparatus for extracting a horizontal line from a marine scene, the method comprising: (a) designating a region of interest in a still image corresponding to every frame of the marine scene movie; and (b) obtaining a horizontal line corresponding to a line alignment in the region of interest of the still image.

Preferably, the still image is contrast-adjusted and Gaussian filtered.

The region of interest may be specified in the first still image among the still images corresponding to each frame.

The coordinate of the ROI may be performed by receiving an upper left point and a lower right point from an input device.

The line alignment calculates a difference between contrast values of two pixels that are vertically symmetric about each pixel in each column in the ROI, and selects maximum points corresponding to pixels having the largest contrast difference in each column. More preferably, the calculation is performed using the maximum points. The column is preferably all columns in the region of interest. The line alignment may be performed by obtaining centers of the maximum points and shifting the maximum points with respect to the center. The step of obtaining the horizontal line is a matrix when the shifted points are xrel (i) and yrel (i), respectively.

Each parameter a, b, which is an eigenvalue of, is obtained, and c is obtained by substituting the parameters a, b and the coordinates (xm, ym) of the center into the basic equation ax + by + c = 0 of a straight line to obtain a horizontal parameter a, It is characterized by obtaining b and c.

As described above, the present invention provides a process of designating a region of interest near a horizontal line in a provided marine scene video and finding a portion having the largest brightness difference by using the luminance difference between pixels corresponding to one row in the region of interest. The horizontal line is extracted by line alignment using the maximum points found by performing each row.

Therefore, the horizontal line extraction result is very stable, and the horizontal line extraction is automated, so there is an advantage that the horizontal line can be easily extracted even in a video scene of several hundred frames.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram showing an embodiment of a horizontal line extraction apparatus according to the present invention, Figure 2 is a flow chart showing an embodiment of a horizontal line extraction method according to the present invention step by step, Figure 3 is a sea life of the horizon is faint 4 illustrates an original image, and FIG. 4 illustrates designation of a region of interest in a first image of a marine scene.

1 to 4, first, the horizontal line extracting apparatus 10 of FIG. 1 includes a still image scanning unit 12, a contrast adjusting unit 14, a Gaussian filter 16, and a region of interest extraction unit 18. And a control unit 20 and connected to an external input device 22 and a display 24.

In the same figure, the still image scanning unit 12 in the horizontal line extracting apparatus 10 scans every frame of the provided marine scene moving image to generate a still image corresponding to every frame, and thus the contrast adjusting unit 14. To provide (S10 of FIG. 2). The still image string constituting the marine scene video must have a consistent file name and serial number so that it can be batch processed later. Therefore, the controller 20 assigns a consistent file name and serial number to the still image columns constituting the marine scene moving image so that the columns of the still image can be batch processed. For example, the controller 20 assigns the same file name to a plurality of scanned still images corresponding to the marine scene video provided through the Gaussian filter 16, and sequentially assigns a serial number to each scanned still image. Thereafter, each scanned still image is batch processed into one file.

The contrast adjusting unit 14 adjusts the contrast of the scanned still image as shown in FIG. 3 provided from the still image scanning unit 12 (S12 of FIG. 2). At this time, if the contrast is adjusted too much, horizontal information may be lost, so appropriate contrast should be given to the scanned still image. The degree of contrast adjustment of the contrast adjuster 14 may be set in advance in the contrast adjuster 14 or may be set according to the control of the controller 20. That is, the contrast adjusting unit 14 adjusts the contrast of the scanned still image provided from the still image scanning unit 12 based on presetting or under the control of the control unit 20.

After adjusting the contrast, a relatively high frequency is applied at the boundary part of the object, that is, at the edge pixel, etc., so that the Gaussian filter 16 is used to reduce noise in an image having a high spatial frequency. Therefore, the Gaussian filter 16 performs Gaussian filtering on the scanned still image whose contrast is provided from the contrast adjusting unit 14, and outputs a Gaussian filtered first still image as shown in FIG. 4 (S14 of FIG. 2). Of course, the Gaussian filter 16 sequentially Gaussian filters the contrast-adjusted still images, which are sequentially provided from the contrast adjuster 14, and provides the same to the controller 20.

The region of interest extractor 18 receives a rectangular coordinate of the region of interest inputted by an input device 22, for example, a mouse drag. The region of interest extractor 18 applies the gaussian filtered first still image through the screen (not shown in the drawing). That is, the controller 20 recognizes the region of interest I in the Gaussian filtered first still image. On the other hand, since a technique for inputting coordinates by the input device 22 such as a mouse has already been generalized, its detailed description will be omitted.

The controller 20 extracts only the ROI designated by the ROI extractor 18 from the Gaussian filtered first still image provided from the Gaussian filter 16 to reduce the amount of calculation (S16 and S18 of FIG. 2). For example, when designating the ROI, the controller 20 receives the upper left point and the lower right coordinate point from the input device 22 through the ROI extractor 18. The ROI extractor 18 processes the area input by the input device as a coordinate point that can be recognized by the controller 20 and transmits the area to the controller 20. For easy understanding, suppose that the specified region of interest is a matrix, and the upper left point is (X1, Y1) and the lower right point is (X2, Y2), the upper left position is (1, 1). , The bottom right point becomes (X2-X1 + 1, Y2-Y1 + 1). The controller 20 extracts a horizontal line from the coordinate point input through the ROI extraction unit 18 through the operation processes S20, S22, and S24 to be described later, and displays the extracted horizontal line through the display 24. . The process calculated by the control unit 20 will be described in more detail with reference to FIGS. 5 to 7 to be described later.

FIG. 5 is a diagram showing the light and dark distribution of one row in the region of interest I designated in FIG. 4, and shows the light and dark distribution from the first column to the last column of any row of the designated ROI matrix.

Referring to FIG. 5, when the image is blurred, it is difficult to clearly distinguish where to determine an edge, such as a contrast distribution appearing at other edges.

Accordingly, the controller 20 may determine the difference between the contrast values of the two pixels that are vertically symmetric about each pixel among the gaussian filtered still images provided from the Gaussian filter 16 in every column in the region of interest I, respectively. By calculating, the maximum points corresponding to the pixels having the largest contrast difference in every column are calculated (S20 of FIG. 2). For example, if there are 1,000 columns in the region of interest I, since the maximum points are calculated for each column, the maximum viscosity is 1,000. The controller 20 calculates a difference between pixel contrast values of -60 pixels above and pixel contrast values of +60 pixels below each pixel for every pixel in the region of interest I, respectively. It can be represented as 1.

ROI (i, j) = I (i + 60, j)-I (i-60, j)

Here, ROI (i, j) means the j th row and the i column of the designated ROI matrix, and I (i, j) means the j th row and the i th column of the input image. In addition, as a result of experimenting with 2048 * 1556, which is the scan image size of a general movie scene, +60 and -60 were used because the best horizontal extraction result was obtained at a difference of about 120 pixels.

FIG. 6 is a diagram illustrating a difference distribution and a maximum point of a light intensity distribution map, and are results of a region of interest according to Equation 1. FIG. Here, suppose the edge of the row is the point where the maximum point among the 60 pixel up and down differences is obtained.

5 and 6 are based on gray levels divided by 250. FIG.

FIG. 7 is a diagram showing horizontal lines found using the maximum points M obtained for each row according to Equation 1, and the controller 20 uses the maximum points M to line-align the lines. The horizontal lines corresponding to the curves are obtained (S22 and S24 in Fig. 2).

For example, in order to find the most suitable line in the scattered maximum points M, the centers of all the maximum points M are obtained, and then the maximum points M are shifted with respect to the center. When the points thus obtained are xrel (i) and yrel (i), the eigenvalues of the following matrix are obtained. The eigenvalues are the parameters a and b, respectively.

Substituting the parameters a, b obtained using the matrix A and the coordinates (xm, ym) of the center into the basic equation ax + by + c = 0 of the straight line to find c, the parameters a, b, c of the horizontal equation can be obtained. have.

If the Gaussian filtered still image provided from the Gaussian filter 16 is not the last still image, the control unit 20 performs step S10 to perform the steps S12 and S14 for every still image up to the last still image. , S16, S18, S20, S22, and S24 are performed (S26). That is, since the ROI does not need to be selected again for the second or the last still image generated in the step S10, the steps S12, S14, S20, S22 and S24 are performed.

In the present invention as described above, when the horizontal line is blurred when the horizontal line is not found by radon or Hough transform, the horizontal line can be obtained.

As described above, the method of the present invention may be implemented as a program and stored in a recording medium (CD-ROM, RAM, ROM, floppy disk, hard disk, magneto-optical disk, etc.) in a computer-readable form.

The present invention described above is not limited to the above-described embodiment and the accompanying drawings, and various substitutions, changes, and conversions can be made without departing from the spirit of the invention. It will be apparent to those who have knowledge.

1 is a block diagram showing an embodiment of a horizontal line extracting apparatus according to the present invention,

2 is a flowchart showing step by step an embodiment of a horizontal line extraction method according to the present invention;

3 is a diagram showing a circle image of a marine life with a blurred horizontal line;

4 is a view showing designation of a region of interest in a first image of a marine scene;

FIG. 5 is a diagram illustrating light and dark distributions of one row in the ROI designated in FIG. 4; FIG.

6 is a view showing a difference distribution and a maximum point of a light intensity distribution map,

7 shows a horizontal line found using the maximum points found for each row.

<Explanation of symbols for the main parts of the drawings>

10: horizontal extraction unit 12: still image scanning unit

14: Contrast adjustment part 16: Gaussian filter

18: ROI extraction unit 20: control unit

22: input device 24: display

Claims (17)

A device for extracting horizontal lines from a marine scene, A region of interest extraction unit configured to extract coordinates of the region of interest selected from the still image corresponding to each frame of the marine scene video; And a control unit for obtaining a horizontal line corresponding to a line alignment in the coordinates of the ROI of the still image. The method of claim 1 The still image is a horizontal line extraction device, characterized in that the contrast is adjusted and Gaussian filtered. The method of claim 2, And said contrast is adjusted based on presetting. The method of claim 2, And said contrast is adjusted under the control of said controller. The method according to any one of claims 1 to 4, And the ROI extracting unit extracts coordinates of the ROI from the first still image of the still images corresponding to each frame. The method according to any one of claims 1 to 4, The line alignment calculates a difference between contrast values of two pixels that are vertically symmetric about each pixel in each column in the ROI, and selects maximum points corresponding to pixels having the largest contrast difference in each column. Calculating, and performing using the maximum points. The method of claim 6, Wherein said column is all columns in said region of interest. The method according to any one of claims 1 to 4, And the control unit assigns the same file name to a plurality of still images corresponding to the sea scene video, and sequentially assigns serial numbers to each still image. The method of claim 1 And a region of interest coordinates extracted by the region of interest extracting unit by performing coordinates of the upper left point and the lower right point input by the input device. A method of extracting horizons from a device that extracts horizons from a marine scene, (a) designating a region of interest in a still image corresponding to every frame of the marine scene video; and (b) obtaining a horizontal line corresponding to a line alignment in the region of interest of the still image. The method of claim 10 The still image is a horizontal line extraction method characterized in that the contrast is adjusted and Gaussian filtered. The method of claim 10 The region of interest is a horizontal line extraction method, characterized in that is specified in the first still image of the still image corresponding to each frame. The method of claim 10 The coordinates of the ROI are performed by receiving input of the upper left point and the lower right point from the input device. The method of claim 10 The line alignment calculates a difference between contrast values of two pixels that are vertically symmetric about each pixel in each column in the ROI, and selects maximum points corresponding to pixels having the largest contrast difference in each column. Calculating, and performing using the maximum points. The method of claim 14, Wherein said column is all columns in said region of interest. The method according to claim 14 or 15, And the line alignment is performed by obtaining the centers of the maximum points and shifting the maximum points with respect to the center. The method of claim 10 Obtaining the horizontal line, The shifted points are referred to as xrel (i) and yrel (i), respectively.

Each parameter a, b, which is an eigenvalue of, is obtained, and c is obtained by substituting the parameters a, b and the coordinates (xm, ym) of the center into the basic equation ax + by + c = 0 of a straight line to obtain a horizontal parameter a, b, c is a horizontal line extraction method characterized in that.

Images