JPH0816779A - Color image processor - Google Patents

Abstract

PURPOSE:To easily and accurately extract an object from an image concerning the image processor. CONSTITUTION:This device is provided with a representative point designating means 1 for respectively designating representative points more than one on the object inside a color image on a screen and representative points more than one on the background, color direction calculating means 2 for calculating a direction vector in the color space of picture elements, and color direction comparing means 3 for comparing the direction vector in the color space of the picture element at an object point with a direction vector in the color space of picture elements at the representative point on the object and the representative point on the background so as to show the attribute of the closest representative point. Further, a representative point selecting means is provided to calculate a distance between the object point and the picture elements at the respective representative points on the object and on the background on the screen and to select the fixed number of representative points in order from the shortest distance.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color image processing apparatus. In particular, it relates to an object extraction process in image processing.

In image processing, when converting a color of a part of an image or combining a part of an image with another image, it is necessary to extract only an object in advance. Therefore, efficient object extraction processing is required.

[0003]

2. Description of the Related Art Conventionally, in object extraction, an operator determines pixels at the boundary between the object and the background (in some cases, enlarges the image).
A method of determining and specifying, and a method of automatically extracting with a certain degree of accuracy by specifying a color are used.

However, in the above-mentioned manual extraction method,
It is necessary for the operator to specify all boundary parts in detail, and with the automatic extraction method, the natural image is originally the same color depending on the lighting condition, but the image has different colors. It is difficult to extract accurately because there are multiple colors.

[0005]

Therefore, manual extraction by an operator has a problem that extraction takes time and labor, and automatic extraction by color designation has a problem that accurate extraction is difficult.

An object of the present invention is to easily and accurately extract an object in an image.

[0007]

[Means for Solving the Problems]

First invention: FIG. 1 shows the principle of the present invention. In a color image processing apparatus, a representative point designating unit 1 for designating one or more representative points on an object and one or more representative points on a background in a color image on a screen, and a direction vector of a pixel in a color space are calculated. The color direction calculation unit 2 compares the direction vector of the pixel of the target point in the color space with the direction vector of the pixel of the representative point on the object and the pixel of the representative point on the background in the color space, and the attribute of the nearest representative point. And a color direction comparing means 3 indicating

With respect to the pixel of each target point on the screen, it is determined whether it belongs to the object or the background according to the judgment of the color direction comparing means 3. Second invention: In contrast to the first invention, the distance between pixels on the screen between the target point and each representative point on the object and the background is calculated, and a fixed number of representative points are arranged in the order of decreasing distance. A representative point selecting means 4 for selecting is provided.

The color direction comparing means 3 compares the direction vector in the color space of the pixel of the target point and the pixel of each representative point selected by the representative point selecting means 4 and indicates the attribute of the closest representative point.

[0010]

FIG. 2 is a diagram for explaining the principle of the present invention. (A) is an example of a display screen, (B) is a diagram illustrating a direction vector in the color space.

When the values of the three primary colors of a pixel in the image are normalized by their absolute values, a direction vector which is a value representing the color is obtained regardless of the brightness. For example, assuming that the three primary colors are red, green, and blue and the pixel value is represented by (R, G, B), (r, g, b) = (R / A, G / A, B / A However, the value (r, g, b) obtained by A ² = R ² + G ² + B ² is a direction vector in the RGB color space. This value is located on the spherical surface shown in FIG. 2B, and if the ratios of the values of R, G, B are the same, they are the same r, g, b, and are the same point. In addition, colors that are close to each other, regardless of brightness, are close to each other on this spherical surface. Therefore, the target point, that is, each pixel in the image, the representative point on the object (object point: x in FIG. 2) and the representative point on the background (background point: x in FIG. 2) on the sphere are defined. By comparing the distances l ₁ and l ₂ , it is possible to determine the closeness of hue. That is,
Whether the target point is closer to the object point or the background point in the color space,
Alternatively, it can be determined whether or not it is on the object by determining whether the object point or the background point is within a certain range from the target point.

By doing so, the operator need only perform a simple operation such as designation of a representative point. Since the direction vector of color is used to judge whether it is an object,
It is possible to appropriately determine a part that is dark due to the lighting condition.

Further, in the second invention, a large number of representative points on the object and the background are designated and selected according to the distance between pixels on the image with the target point to be used. Even if it is, it can be appropriately determined.

[0014]

FIG. 3 is a block diagram of an embodiment of the present invention. FIG.
Those having the same functions as those are indicated by the same reference numerals. Embodiments of the present invention will be described below with reference to the drawings.

The image processing apparatus includes a memory 10 and a processor 50.
Image memory 60, display device 61, mouse 62, keyboard 63, disk device 70, and other input / output device 80.
And the image processing program 20 and the hardware connected by the system bus 90. In this embodiment, a program 40 for extracting an object is added to this. The object extraction program 40 includes a representative point designation routine 1, a color direction calculation routine 2, a representative point selection routine 4, a color direction comparison routine 3 and a main routine 0. Further, there is a representative point input routine 31 as a related routine.

FIG. 4 is a flow chart of this embodiment.
The operation procedure will be described with reference to FIG. S1: First, the operator specifies the background and some points on the object. In this embodiment, an operator designates a representative point by drawing an appropriate curve on the object and on the background, respectively. This may be configured as the representative point input routine 31 using a known technique such as line drawing with a mouse. The extraction processing range may be specified separately, but if the background is specified by a closed curve, it can be considered that the extraction processing range is specified at the same time. S2: The representative point designation routine 1 sets points obtained by appropriately dividing each curve and sets them as representative points (see FIG. 5A). S3: The color direction calculation routine 2 calculates the direction vector of these designated representative points in the RGB space. The result is held as table 5. S4: Next, the main routine 0 sequentially extracts pixels within the extraction processing range and performs the following processing. S5: The representative point selection routine 4 calculates the distance to each representative point based on the geometrical position on the screen. Among them, a fixed number of points (3 points in the example in the figure) are selected in descending order. S6: The color direction calculation routine 2 calculates the direction vector of the pixel. S7: The color direction comparison routine 3 regards the value (r, g, b) of the direction vector as the coordinate of the three-dimensional space, and the pixel and the direction vector of each of the four representative points selected by the representative point selection routine. The distance to the value of is calculated, the smallest (close) representative point is examined, and if it is the object point, the target pixel is the object, and if not, it is the background. S8: For all the pixels within the extraction processing range, S4 to
The process of S7 is performed and the process ends.

FIG. 5 is an explanatory diagram of the processing of the embodiment, which is an example of determination for one point P within the extraction processing range. Figure 5
(A) shows the positional relationship between the curve on the object designated by the operator, the closed curve on the background, the object point and background point set therefrom, and the point P to be determined. First, three object points and three background points close to the point P to be determined are taken. This is to deal with the case where the object or the background is not a single color, and the number is determined in advance according to the number of colors of the target image.

FIG. 3B shows three object points o1, o2, o3 and a background point b with respect to the point P ₁ obtained by the above process.
It shows 1, b2, b3. Also, three object points o4, o5, o6 and background points b4, b5, b6 for the point P ₂ are shown.

Next, the point P is defined as the object points o1, o2, o3 and the background point.
By comparing b1, b2, b3 with the direction vector, it is determined whether it is an object or a background. When the direction vector of the representative point x in the RGB space is (rx, gx, bx), the difference in the direction vector between the point P and the representative point x is represented by coordinates (rp, gp, b).
p) and the coordinates (rx, gx, bx). This distance may be calculated as the geometrical distance (square of) of three-dimensional space (rx-rp) ² + (gx-gp) ² + (bx-bp) ² , but an accurate value is not always necessary. Since it does not exist, it may be simply │rx−rp│ + │gx−gp│ + │bx−bp |.

In this way, the distances in the RGB space from the plurality of representative points are calculated. The representative point having the smallest distance among them, that is, the direction of the direction vector is close (the RGB ratio is close, and the color is close) is determined. If it is an object point, the point P is determined to be an object, and if it is a background point, it is determined to be a background. Or, especially when the number of representative points is large,
Depending on the number of representative points whose distance in the RGB space is within a certain range, it may be determined that they belong to the larger one.

In the example of the figure, the representative point having the direction vector closest to the direction vector of the point P ₁ should be the object point o 1. Therefore, the object is determined. If the target point is in the yellow range slightly below the point P _{1 in} the figure, the representative point having the closest direction vector should be the object point o 2. Therefore, it is also determined as an object. For the point P ₂ , background points b4, b5,
One of b6 is the closest direction vector. In this way, an appropriate determination can be made even when the object or background is not a single color.

When the RGB value of the point P is very small (that is, low brightness), the error in the value of the direction vector can be large, and it is more accurate to exclude it as undecidable. In that case, it is determined in combination with another determination process or by an operator's instruction.

The above-mentioned processing is performed for all the pixels within the extraction processing range to perform noise removal processing such as removal of isolated points from the data separated into the object and the background, and the extraction processing of the object portion is completed.

[0024]

As described above, the image processing apparatus according to the present invention can extract an object in an image by a simple operation of designating a representative point. In addition, since the difference in brightness can be removed and the same color / approximate color portion can be accurately identified by using the information on the direction of the pixel in the color space, it is possible to extract the object with considerably high precision.

[Brief description of drawings]

[Figure 1] Principle diagram

[Figure 2] Principle explanation diagram

FIG. 3 is a configuration diagram of an embodiment

FIG. 4 is a flowchart of an embodiment.

FIG. 5 is an explanatory diagram of the processing of the embodiment.

[Description of Reference Signs] 1 representative point specifying means (representative point specifying routine) 2 color direction calculating means (color direction calculating routine) 3 color direction comparing means (color direction comparing routine) 4 representative point selecting means (representative point selecting routine) 5 Representative point direction vector table 10 Memory 20 Image processing program 30 Input / output processing program 40 Object extraction program 50 Processor 60 Image memory 61 Display device 62 Mouse 63 Keyboard 70 Disk device 80 Other input / output device 90 System bus

Claims (2)

[Claims] 1. In a color image processing apparatus, a representative point designating means (1) for designating one or more representative points on an object and representative points on a background in a color image on a screen, and a pixel color. A color direction calculation means (2) for calculating a direction vector in the space, a direction vector in the color space of the pixel of the target point, a representative point on the object, and a direction vector in the color space of the pixel of the representative point on the background. A color direction comparing means (3) indicating the attribute of the closest representative point is provided, and the pixel of each target point on the screen is determined to belong to the object or on the background according to the judgment of the color direction comparing means (3). A color image processing apparatus, characterized in that it determines whether or not it belongs to. 2. A representative point selecting means for calculating a distance between pixels on a screen between a target point and each representative point on an object and a background, and selecting a fixed number of representative points in ascending order of distance ( 4) is provided, and the color direction comparison means (3) compares the direction vector in the color space of the pixel of the target point and the pixel of each representative point selected by the representative point selection means (4), and the closest representative The color image processing apparatus according to claim 2, wherein the color image processing apparatus indicates a point attribute.