JP5227226B2 - Image processing apparatus and image processing program - Google Patents

Description

  The present invention relates to an image processing technique for generating a key image in which a background and a foreground of an input image are distinguished, and more particularly to an apparatus and a program for generating a key image based on a difference between an input image and a background image.

  Conventionally, research and development of a method for detecting only a moving object from a moving image (input image) has been progressed, and for example, a chroma key method, a method using infrared light, and a background subtraction method are known. . In the chroma key method, a subject (animal body) is imaged in a studio or the like with a specific background such as a blue back, and a background image is generated by discriminating a specific color from the image (input image) obtained by the imaging. Then, the moving object is detected from the input image assuming that the moving object other than the background image is an moving object. In the technique using infrared light, a distance image is generated by irradiating a subject with infrared light, and a moving object is detected from an input image using the distance image. The background subtraction method assumes that an input image includes a background image and an image of a moving object, generates a background image from the input image, and obtains a difference between the input image and the background image. It detects an animal body.

  As an example using the chroma key method, a chroma key device that generates a key image by specifying a background color range and determining whether or not it belongs to the specified background color for each pixel in the input image is known. (See Patent Document 1 and Patent Document 2).

  As an example of a technique using infrared light, the infrared light round trip time is calculated by irradiating the subject with infrared light, and distance information for each pixel is extracted based on the round trip delay time. An apparatus that generates a key image based on information is known (see Patent Document 3).

  Furthermore, as an example using the background difference method, a background image is generated from a past input image, and based on the difference between the background image and the input image, an area whose pixel value change is less than a threshold is regarded as the background. Background difference devices (difference key devices) that generate key images are known (see Patent Literature 4, Patent Literature 5, and Patent Literature 6).

  Here, the key image refers to an image in which the background and the foreground of the input image are distinguished from each other, for example, an image in which the background and the foreground are distinguished by binary values, or an image in which multiple values are distinguished. In an image processing apparatus that generates a key image by analyzing an input image, it is important to generate a highly accurate key image that can correctly determine the background and foreground of the input image.

Japanese Patent No. 4004863 JP 2007-142726 A Japanese Patent No. 4031306 Japanese Patent No. 3228638 Japanese Patent No. 3123587 Japanese Patent No. 28828380

  In the chroma key devices of Patent Documents 1 and 2 described above, a cloth, board, wall or the like having a uniform color such as green or blue called a chroma key back or chroma key screen is used as a background, or a lawn or the like. It was necessary to use various natural or artificial objects as the background. For this reason, it is necessary to use a background having a preset color, and there is a problem that image processing is restricted.

  Further, in the device using the round trip delay time by infrared light described in Patent Document 3 described above, when the projected infrared light reaches the subject, that is, when a subject that can be irradiated with infrared light is treated as a foreground, This method can be applied. However, since the subject exists in the distance, there is a problem that it is difficult to apply this method when the projected infrared light does not reach the subject, that is, when it is desired to treat the far subject as the foreground.

  Furthermore, in the background difference device described in Patent Documents 4 and 5 described above, the update of the background image is controlled based on the spatially local feature of the input image to generate a key image. However, there is a possibility that an unintended pattern may appear in the generated key image, and there is a problem that a highly accurate key image that can correctly determine the background and the foreground of the input image is not generated. This is because, for example, even when ambient light fluctuations such as illumination fluctuations occur and the pixel value of the input image changes, a key image is once generated, and then a background corresponding to the changed pixel value. This is because the image is updated. That is, an unintended pattern appears in the key image due to the changed pixel value, and then a background image for excluding the unintended pattern from the key image is generated.

  Further, the background difference device of Patent Document 6 described above uses a filter for removing unintended noise included in the key image. However, this filter performs processing on the generated key image to reduce noise in the key image. Therefore, this filter is used after executing the background subtraction method, and is not used in the process. That is, it is not a filter used for the background subtraction method.

  Therefore, the present invention has been made to solve the above-described problems, and the object thereof is a case where an ambient light fluctuation such as an illumination fluctuation occurs when generating a key image using the background subtraction method. Another object of the present invention is to provide an image processing apparatus and an image processing program in which the variation does not appear in the key image.

  In order to solve the above-described problems, the present invention has the following configuration.

  The image processing device according to claim 1, wherein a key image that distinguishes a background and a foreground of an input image is generated by background difference processing that subtracts a background image from the input image, and the image processing device at a plurality of past points in time is generated. A background image generating means for generating a background image from the input image, a background image storing means for storing the background image generated by the background image generating means, a background image stored by the input image and the background image storing means, Based on the first threshold, and based on the comparison result, background update control means for controlling the timing at which the background image is generated by the background image generation means, the input image and the background image storage means The stored background image is compared based on the second threshold value, and the key image generating means for generating the key image is provided.

  According to such a configuration, the image processing apparatus synthesizes an image of only the background that does not include the foreground including the moving object from the past input image in the background image generation unit, and stores it in the background image storage unit. Further, the image processing apparatus generates a key image by comparing the background image stored in the background image storage unit with the input image in the key image generation unit. Further, the image processing apparatus compares the background image stored in the background image storage unit with the input image in the background update control unit, and controls the timing at which the background image generation unit generates the background image.

  The image processing apparatus according to claim 2 is the image processing apparatus according to claim 1, wherein the background image generation unit stores the input image at a plurality of past time points, and the past image storage. Filter processing in the time direction for each pixel position of the input image stored by the means, and based on the comparison result by the time filter means for generating a background image and the background update control means, the time filter means performs the filtering process. And a control means for controlling the timing at which the background image is generated.

  According to this configuration, the image processing apparatus performs the time-direction filtering process on each pixel of the past input image stored in the past image storage unit according to the timing control by the control unit in the time filter unit. A background image is generated by obtaining a representative value.

  The image processing apparatus according to claim 3 is the image processing apparatus according to claim 2, wherein the control unit of the background image generation unit uses the past image storage unit based on the comparison result by the background update control unit. The timing at which the input image is stored and the timing at which the background image is generated by performing the filtering process by the time filter means are controlled.

  According to this configuration, the image processing apparatus sequentially stores input images at a plurality of past points in the past image storage unit according to the timing control by the control unit. The image processing apparatus obtains a representative value of the pixel by performing time-direction filtering on each pixel of the past input image stored in the past image storage unit in the time filter unit according to the timing control by the control unit. As a result, a background image is generated.

  The image processing device according to claim 4 is the image processing device according to any one of claims 1 to 3, wherein a threshold value is set for setting the first threshold value and the second threshold value. It was set as the structure provided with a means.

  According to such a configuration, the image processing apparatus uses the threshold value setting unit to set the first threshold value used by the background update control unit and the second threshold value used by the key image generation unit. Set and change according to operation.

  Although the present invention has been described as an image processing apparatus, the present invention can be realized as a program substantially corresponding to these, and the present invention includes an image processing program. That is, the image processing program according to claim 5 is an image processing program for generating a key image that distinguishes a background and a foreground of an input image by a background difference process that subtracts the background image from the input image. A step of generating a background image from the input images at a plurality of past points, a step of storing the generated background image, and comparing the input image and the stored background image based on a first threshold, Controlling the timing for generating the background image based on the comparison result, comparing the input image with the stored background image based on a second threshold, and generating the key image; Is configured to be executed.

  As described above, according to the present invention, even when ambient light fluctuation such as illumination fluctuation occurs and the background changes, the background image is updated before the fluctuation is reflected in the key image, A background image reflecting the change was generated. Therefore, since the key image is generated based on the background image in which the variation is reflected, the variation does not appear in the key image. Therefore, it is possible to generate a key image that can accurately determine the background and foreground of the input image.

It is a block diagram which shows the structure of the background difference apparatus by embodiment of this invention. It is a block diagram which shows the structure of a background image generation means. It is a flowchart explaining the outline of operation | movement in a background difference apparatus.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[Configuration of background difference device]
First, the configuration of the background difference device will be described. FIG. 1 is a block diagram showing a configuration of a background difference device according to an embodiment of the present invention. The background difference device (image processing device) 1 includes a background image generation unit 10, a background image storage unit 20, a background update control unit 30, a key image generation unit 40, and a threshold setting unit 50.

The threshold setting means 50 is a man-machine interface, and sets a _first threshold θ ₁ and a second threshold θ ₂ . The first threshold value θ ₁ set by the threshold value setting means 50 is output to the background update control means 30, and the second threshold value θ ₂ is output to the key image generation means 40. For example, when the user (operator) who operates the rotary knob or the translational slider increases or decreases the first threshold value θ ₁ and the second threshold value θ ₂ , the threshold value setting means 50 is configured to change the first threshold value θ 1. _First and second threshold values θ ₂ are set. The threshold value may be set by increasing / decreasing a threshold value or specifying a numerical value by a user's key operation. Here, the background update control means 30 using the first threshold value θ ₁ operates more sensitively to variations in the input image than the key image generation means 40 using the second threshold value θ ₂ . θ ₁ <θ ₂ is set.

The background update control means 30 inputs the input image, reads out and inputs the background image from the background image storage means 20, and inputs the first threshold θ ₁ from the threshold setting means 50. Then, the background update control unit 30 generates the determination result f based on the pixel value I (t, x, y) of the current input image, the pixel value B (x, y) of the background image, and the threshold value θ _1. . The determination result f generated by the background update control unit 30 is output to the background image generation unit 10. Here, the pixel value of the input image at time t and the pixel position (x, y) (x is the horizontal coordinate, y is the vertical coordinate) is I (t, x, y), and the pixel value of the background image is B (x , Y).

Specifically, the background update control means 30 treats the first threshold value θ ₁ as a threshold value for the entire image, and the pixel value I (t, x, y) of the input image and the pixel value B (x, y) of the background image. ) Is calculated for each pixel, and it is determined whether or not the sum of the absolute values of all the pixels exceeds the _first threshold θ1, and the determination result f is obtained. According to the following formula, when it is determined that the sum of absolute values exceeds the _first threshold value θ ₁ , f = 1 is set, and when it is determined that the sum does not exceed the first threshold θ ₁ , f = 0 is set.

The background update control means 30 treats the first threshold θ ₁ as a threshold for one pixel, and the pixel value I (t, x, y) of the input image and the pixel value B (x, y) of the background image the absolute value of the difference value calculated for each pixel between, it is determined whether the absolute value of each pixel exceeds a first threshold value theta _1, calculates the number of pixels that exceeds the first threshold value theta ₁ To do. Then, it may be determined whether or not the number exceeds a certain ratio φ, and the determination result f may be obtained. According to the following equation, when it is determined that the number exceeds a certain ratio φ, f = 1 is set, and when it is determined that the number does not exceed, the setting is set to f = 0.

The background update control means 30 treats the first threshold value θ ₁ as a threshold value for the entire image, and calculates the pixel value I (t, x, y) of the input image and the pixel value B (x, y) of the background image. Alternatively, the absolute value of the difference value may be calculated for each pixel, and the sum of the absolute values of all the pixels may be divided by the first threshold θ ₁ and the division result may be obtained as the determination result f. The determination result f is obtained by the following equation.

As described above, the determination result f generated by the background update control unit 30 is a numerical value indicating the degree of difference between the input image and the background image, and a larger value is set as the difference is larger. Further, since the determination result f is generated based on the first threshold value θ ₁ having a value smaller than the second threshold value θ ₂ , if there is a change in the background, the change in the background is the key image. The determination result f is reflected faster in time than the key image generated based on the second threshold value θ ₂ in the generation means 40.

  The background image generation means 10 inputs an input image, receives a determination result f from the background update control means 30, and stores the input image in a memory at a predetermined time interval S based on the determination result f. Further, based on the determination result f, a predetermined amount of input images (past images) are read from the memory, and temporal filter processing is performed on the past input image groups to generate a background image. The background image generated by the background image generation unit 10 is output to and stored in the background image storage unit 20.

  The pixel value I (t, x, y) of the input image is a scalar quantity representing luminance information in a monochrome camera (including an infrared camera), and a color (or wavelength, in a color or multi-wavelength camera). This is a vector quantity representing luminance information for each (spectral characteristic).

  On the other hand, the pixel value B (x, y) of the background image may represent a scalar quantity or vector quantity of the same dimension as the pixel value I (t, x, y) of the input image, or a color image May be a scalar quantity obtained by converting the color to monochrome. Further, it may be a higher-dimensional vector amount or tensor amount for storing the dispersion value for each pixel. Hereinafter, the pixel value B (x, y) of the background image will be described as a scalar quantity or a vector quantity having the same dimension as the pixel value I (t, x, y) of the input image.

  FIG. 2 is a block diagram showing a configuration of the background image generating means 10 shown in FIG. The background image generation unit 10 includes a past image storage unit 11, a time filter unit 12, and a control unit 13.

  The control unit 13 receives the determination result f from the background update control unit 30, sets a time interval S based on the determination result f, and generates a storage timing signal for each time interval S. The storage timing signal generated by the control unit 13 is output to the past image storage unit 11. Thus, the past image storage unit 11 stores the input image in the image memory at the time interval S. Details will be described later. The time interval S is a time difference between the timing when the previous input image is stored in the past image storage unit 11 and the timing when the current input image is stored in the past image storage unit 11.

例えば、減少関数ｇ（ｆ）として、以下のステップ関数を用いるようにしてもよい。

ここで、Ｓ_０＞Ｓ_１＞０とする。また、以下の折れ線関数を用いるようにしてもよい。

Specifically, the control means 13 sets the time interval S so that the time interval S becomes longer when the value of the determination result f is small, and the time interval S becomes shorter when the value of the determination result f is large. . Further, the background update control means 30 may set the time interval S using a decreasing function g (f) whose value range exceeds 0 by the following equation.

For example, the following step function may be used as the decrease function g (f).

Here, S ₀ > S ₁ > 0. Further, the following line function may be used.

  The control means 13 generates an operation signal or a non-operation signal based on the determination result f in addition to the setting of the time interval S and the generation of the storage timing signal described above. The operation signal or non-operation signal generated by the control unit 13 is output to the time filter unit 12. As a result, when the operation signal is input, the time filter unit 12 reads a predetermined amount of input image group from the past image storage unit 11, performs a filtering process in the time direction, and generates a background image. Details will be described later.

  Specifically, the control unit 13 compares the value of the determination result f with a predetermined constant value, and when the value of the determination result f exceeds the predetermined value (for example, when f> 0.5). In addition, an operation signal for operating the time filter unit 12 is generated and output to the time filter unit 12. Further, the control means 13 generates a non-operation signal for not operating the time filter means 12 when the value of the determination result f does not exceed a certain value (for example, when f ≦ 0.5). Output to means 12.

The past image storage unit 11 has an image memory, inputs an input image, inputs a storage timing signal for each time interval S from the control unit 13, and inputs the input image at a timing when the storage timing signal is input. Store in image memory. In the image memory of the past image storage means 11, pixel values I (t ′, x, y) (where t ′ <current time t) of past input images are for N time points (N is an integer of 2 or more). Remembered. Hereinafter, the time of the past input image stored in the past image storage unit _{11, t 1, t 2,} ···, t N ( _{where, t> t 1> t 2} >···> t N) and To do. In the case where the past image storage means 11 has already stored the input images for N times, in order to store a new input image in the image memory at the timing when the storage timing signal is input, the past image storage means 11 stores Of the input images (N past input images), the oldest input image is overwritten. In this case, the new input image is stored as the pixel value I (t ₁ , x, y) of the input image. Thereby, the past image storage unit 11 can always store the pixel values (t ′, x, y) of the past input image for the N time points in the image memory. The past image storage unit 11 may also store the pixel value of the current input image.

  The time filter unit 12 receives an operation signal or a non-operation signal from the control unit 13, and reads a maximum of N past input image groups from the past image storage unit 11 as past images at the timing when the operation signal is input. Then, the time filter unit 12 performs a filtering process in the time direction based on the past images as the input image group to generate a background image. The background image generated by the time filter unit 12 is stored in the image memory of the background image storage unit 20.

  When the non-operation signal is input from the control unit 13, the time filter unit 12 does not perform input image read processing, filter processing, or the like, and at the timing when the operation signal is input, the input image group from the past image storage unit 11. Is filtered and a background image is generated and written into the background image storage means 20.

The time filter unit 12 may perform processing using all of the maximum N past input image groups read from the past image storage unit 11 or may perform processing using a part thereof. May be. Of the time of the past input image group read from the past image storage unit 11, when the set of times of the input image on which the time filter unit 12 performs the filtering process is represented by T, typically, it is expressed by the following expression. .

尚、入力画像がカラー画像の場合、時間フィルタ手段１２は、色成分毎に中央値演算処理を行ってもよいし、入力画像をモノクローム画像に変換してから中央値演算処理を行い、モノクロームにおける背景画像の画素値Ｂ（ｘ，ｙ）を生成するようにしてもよい。 In the filtering process, the time filter unit 12 performs order statistics in the time direction for each pixel position on the past input image group read from the past image storage unit 11, for example, and generates a background image B (x, y) The pixel value at the pixel position in is determined. Specifically, the time filter unit 12 functions as an order statistical filter, and performs a median operation by the following formula as order statistics to generate a pixel value B (x, y) of the background image. med represents the median calculation.

When the input image is a color image, the time filter unit 12 may perform median calculation processing for each color component, or convert the input image to a monochrome image and then perform median calculation processing. The pixel value B (x, y) of the background image may be generated.

  Thus, according to the control means 13 of the background image generation means 10, the time interval S for storing the input image is set based on the determination result f. As a result, when the background changes rapidly in the input image, the determination result f indicating the degree of difference between the input image and the background image is a large value, and the time interval S is shortened. 11 stores an input image for each finely chopped time. On the other hand, when there is almost no background fluctuation, the determination result f becomes a small value and the time interval S becomes long. Therefore, the past image storage means 11 stores an input image for each time that is not so finely engraved. Is done. Accordingly, the filtering process is performed by the time filter means 12 based on the input image group of the time length reflecting the background fluctuation speed, so that a background image reflecting the background fluctuation speed is generated, thereby generating a key image. Is done. That is, the change does not appear in the key image, and it is possible to generate a key image that can accurately determine the background and the foreground of the input image.

  The control means 13 of the background image generation means 10 generates a storage timing signal at a preset fixed time interval S (for example, a time length of 100 frame intervals), and the past image storage means 11 The input image may be stored at the time interval S.

  Further, the control means 13 of the background image generation means 10 generates an operation signal for operating the time filter means 12 or a non-operation signal for not operating based on the determination result f, and the time filter means 12. Output to. As a result, when the background changes rapidly in the input image, the determination result f indicating the degree of difference between the input image and the background image becomes a large value, and an operation signal is generated when the value exceeds a certain value. Then, the time filter unit 12 operates to generate a background image reflecting the background fluctuation. Therefore, a key image is generated from the background image reflecting the background variation. That is, the change does not appear in the key image, and it is possible to generate a key image that can accurately determine the background and the foreground of the input image. On the other hand, when there is almost no change in the background, the determination result f becomes a small value and a non-operation signal is generated. Therefore, the time filter means 12 does not operate, and the key is determined by the already stored background image. An image is generated. In this case, since the background hardly changes, the accuracy of the generated key image does not decrease. Further, when the accuracy of the key image is not affected, the time filter unit 12 is not operated, so that unnecessary operation is eliminated and efficient processing can be realized.

  Although the background image generation unit 10 includes the control unit 13, the background update control unit 30 may include the control unit 13. In this case, the background image generation means 10 includes the past image storage means 11 and the time filter means 12, and the control means 13 of the background update control means 30 sends the storage timing signal to the past image storage means 11 of the background image generation means 10. The operation signal or the non-operation signal is output to the time filter unit 12 of the background image generation unit 10.

  Returning to FIG. 1, the background image storage means 20 has an image memory, inputs the background image B (x, y) generated by the background image generation means 10 and stores it in the image memory.

The key image generation unit 40 inputs the input image, reads out and inputs the background image from the background image storage unit 20, and inputs the second threshold θ ₂ from the threshold setting unit 50. Then, the key image generation unit 40 generates the key image K (t, t, y) based on the pixel value I (t, x, y) of the current input image, the pixel value B (x, y) of the background image, and the threshold value θ ₂ . x, y) is generated and output.

  The key image K (t, x, y) may be binary or multi-valued. In the case of binary, for example, when the key image K (t, x, y) = 1, the image at the pixel position (x, y) of the input image I (t, x, y) is in the foreground, When the key image K (t, x, y) = 0, the image at the pixel position (x, y) of the input image I (t, x, y) is defined as being in the background. In the case of multivalue, for example, the key image K (t, x, y) ∈ [0, 1] is set, and the larger the value of the key image K (t, x, y), the larger the value at the pixel position (x, y). An image is defined as having a high possibility of being in the foreground, and a smaller value indicating a higher possibility of being in the background.

Specifically, the key image generation means 40 generates a binary key image K (t, x, y) by the following formula.

ここで、第２の閾値θ_２は、下側閾値θ_２ａ及び上側閾値θ_２ｂであるものとする。尚、θ_２ａ＜θ_２ｂである。 Further, the key image generation means 40 generates a multi-value key image K (t, x, y) by the following equation.

Here, the second threshold θ ₂ is assumed to be the lower threshold θ _2a and the upper threshold θ _2b . Note that θ _2a <θ _2b .

In this way, the key image generation unit 40 determines the key image K based on the pixel value I (t, x, y) of the current input image, the pixel value B (x, y) of the background image, and the threshold value θ _2. (T, x, y) is generated and output. Since the key image K (t, x, y) is generated based on the second threshold θ ₂ , if there is a change in the background in the input image, the change in the background is first controlled by the background update control. The determination result f generated based on the first threshold value θ ₁ in the means 30 and the background image generated based on the determination result f in the background image generation means 10 are reflected. Since the key image generation unit 40 generates the key image K (t, x, y) based on the background image reflecting the background change, the background change is generated by the key image K (t, x, y). ) Will not appear. Therefore, it is possible to generate a key image K (t, x, y) that can accurately determine the background and foreground of the input image. In addition, when there is a change in the background of the input image, the change is first reflected in the background image, and then the key image K (t, x, y) is generated based on the background image. In the threshold setting means 50, the first threshold θ ₁ and the second threshold θ ₂ are set. Further, the control means 13 sets a constant value which is an operation signal generation condition for operating the time filter means 12 to generate a background image.

[Operation]
Next, the operation of the background difference device 1 shown in FIG. 1 will be described. FIG. 3 is a flowchart for explaining the outline of the operation in the background difference device 1. First, the threshold setting unit 50 of the background difference device 1 sets the first threshold value theta ₁ and the second threshold value theta ₂ (step S301). The first threshold value θ ₁ and the second threshold value θ ₂ set here are θ ₁ <θ ₂ . This is because when the background of the input image changes, before the change is reflected in the key image, a background image that reflects the change is generated so that the change does not affect the key image. is there. That is, the background update control means 30 using the first threshold value θ ₁ uses the second threshold value θ ₂ to generate a key image before the key image generation means 40 generates the key image. It is for generating.

The control unit 13 of the background image generation unit 10 generates the determination result f based on the input image, the background image, and the first threshold value θ ₁ (step S302). In the determination result f generated here, when the background of the input image changes, the change is reflected before the key image generated in step S305 described later.

  The past image storage unit 11 of the background image generation unit 10 stores the input image according to the storage timing signal based on the determination result f generated by the control unit 13 (step S303). Further, the time filter unit 12 of the background image generation unit 10 reads the input image group as a past image from the past image storage unit 11 in accordance with the operation signal based on the determination result f generated by the control unit 13, and performs time direction filter processing. To generate a background image (step S304). Similar to the determination result f, when the background of the input image changes, the background image generated here reflects the change prior to the key image generated in step S305 described later.

The key image generation means 40 generates a key image based on the input image, the background image, and the second threshold value θ ₂ (a value larger than the first threshold value θ ₁ ) (step S305).

  As described above, according to the background difference device 1 according to the embodiment of the present invention, when the background of the input image changes due to a change in ambient light such as a change in illumination, first, the background update control unit 30 performs the background image generation unit. 10, a background image reflecting the fluctuation is generated. After that, the key image generation means 40 generates a key image using the background image reflecting the fluctuation. As a result, the fluctuation of the input image is absorbed by the background image, so that the key image is not affected. Therefore, no noise is generated in the key image due to the fluctuation of the background of the input image, and the generation of the noise can be prevented in advance. That is, it is possible to generate a key image that accurately discriminates the background and foreground of the input image.

  Further, according to the background difference device 1 according to the embodiment of the present invention, the past image storage unit 11 of the background image generation unit 10 stores past input images for a predetermined time. Thereby, in order to generate a background image using a past input image group, a time filter can be used as a filter.

Further, according to the background difference device 1 according to the embodiment of the present invention, the threshold setting unit 50 includes the first threshold θ ₁ used by the background update control unit 30 and the first threshold used by the key image generation unit 40. The threshold value θ ₂ of ₂ is set so as to satisfy θ ₁ <θ ₂ . Thus, after the background update control unit 30 causes the background image generation unit 10 to generate a background image in which the variation of the input image is reflected, the key image generation unit 40 uses the background image in which the variation is reflected. A key image can be generated. Therefore, no noise is generated in the key image due to the fluctuation of the background of the input image, and the generation of the noise can be prevented in advance. That is, it is possible to generate a key image that accurately discriminates the background and foreground of the input image. Further, a desired key image can be generated by adjusting the first threshold value θ ₁ and the second threshold value θ ₂ .

  The background difference device 1 is configured by a computer including a volatile storage medium such as a CPU and a RAM, a non-volatile storage medium such as a ROM, an interface, and the like. The functions of the background image generation means 10, background image storage means 20, background update control means 30, key image generation means 40, and threshold value setting means 50 provided in the background difference device 1 are stored in the CPU with a program describing these functions. Each is realized by executing. These programs can also be stored and distributed in a storage medium such as a magnetic disk (floppy (registered trademark) disk, hard disk, etc.), optical disk (CD-ROM, DVD, etc.), semiconductor memory, or the like.

DESCRIPTION OF SYMBOLS 1 Background difference apparatus 10 Background image generation means 11 Past image storage means 12 Time filter means 13 Control means 20 Background image storage means 30 Background update control means 40 Key image generation means 50 Threshold setting means

Claims (5)

In an image processing apparatus that generates a key image that distinguishes the background and foreground of an input image by background difference processing that subtracts the background image from the input image,
Background image generation means for generating a background image from the input images of a plurality of past points;
Background image storage means for storing the background image generated by the background image generation means;
The input image and the background image stored by the background image storage unit are compared based on a first threshold value, and the timing at which the background image generation unit generates the background image is controlled based on the comparison result. Background update control means;
An image processing apparatus comprising: a key image generation unit configured to compare the input image with a background image stored by the background image storage unit based on a second threshold value and generate the key image. The background image generating means
Past image storage means for storing the input images of a plurality of past points;
Time filter means for performing a filtering process in the time direction for each pixel position of the input image stored by the past image storage means, and generating a background image;
The control unit according to claim 1, further comprising: a control unit configured to control a timing at which a background image is generated by performing a filtering process by the time filter unit based on a comparison result by the background update control unit. Image processing device. The control means of the background image generation means is
Controlling the timing at which the input image is stored by the past image storage unit and the timing at which the background image is generated by performing the filtering process by the time filter unit, based on the comparison result by the background update control unit. The image processing apparatus according to claim 2.   The image processing apparatus according to any one of claims 1 to 3, further comprising a threshold setting unit configured to set the first threshold and the second threshold. An image processing program for generating a key image that distinguishes a background and a foreground of an input image by background difference processing that subtracts a background image from the input image, and
Generating a background image from the input images of a plurality of past points;
Storing the generated background image;
Comparing the input image and the stored background image based on a first threshold, and controlling the timing of generating the background image based on the comparison result;
A step of comparing the input image with the stored background image based on a second threshold value and generating the key image.

Images