JP2011171795A - Noise reduction device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce residual image appearing in the contour shape of a mobile object which is an object on a screen when frame cyclic noise reduction is performed. <P>SOLUTION: A motion determination part 104 selects the an alternative pixel of a larger inter-frame difference value of two alternative pixels which are separated by a plurality of pixels from a pixel of interest in left and right directions as the pixel of a motion determination object. The motion determination part 104 supplies the value of a cyclic coefficient (1-K) to a multiplication part 105 larger as the inter-frame difference value of the selected alternative pixel is larger and supplies the value of a cyclic coefficient K to a multiplication part 106 smaller. Thus, the residua image is suppressed by making the ratio of the multiplied signals of a previous frame from the multiplication part 106 added in an addition part 107 to the multiplied signals of the present frame from the multiplication part 105 smaller when the motion of the pixel of interest is larger, and noise is reduced by making the ratio larger when the motion of the pixel of interest is smaller. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a noise reduction apparatus, and more particularly to an apparatus for reducing noise in a video signal.

  A three-dimensional noise reduction device is known as a technique for removing random noise in a video signal. This three-dimensional noise reduction device pays attention to the fact that the noise component in the video signal has no correlation between frames, and reduces the noise by obtaining an average between a plurality of frames of the input video signal.

  In this type of three-dimensional noise reduction device, the difference between the current input video signal and the video signal of the previous frame from the frame memory is taken, and the signal obtained by multiplying the difference by the cyclic coefficient is added to or subtracted from the input video signal. A frame recursive noise reduction apparatus is configured to output a video signal with reduced noise and supply it to the frame memory.

  However, this noise reduction device can reduce noise in a still image portion in an input video signal, but an afterimage remains in a moving image portion when an average is simply obtained. Therefore, conventionally, motion determination is performed based on the inter-frame difference value between the input video signal and the output video signal of the previous frame, the after-image is suppressed by reducing the amount of circulation for the moving image portion, and the amount of circulation is reduced for the still image portion. Increase to reduce noise.

  Here, the inter-frame difference value includes a difference due to motion and a difference due to noise. For this reason, in the noise reduction device that controls the amount of circulation by the inter-frame difference value, if the visibility of the motion is improved, the ratio of false detection of the difference due to the noise increases as a result of which the noise is conspicuous. There is a trade-off relationship that the afterimage of the motion becomes conspicuous as a result of an increase in the rate of erroneously detecting the difference due to motion as noise.

  For this reason, a technique for distinguishing between noise and movement is important. Therefore, in the noise reduction device described in Patent Document 1, first, in an area of 15 pixels centered on the target pixel as shown in FIG. 20, the number of moving pixels of the inter-frame difference value and the number of still pixels of the inter-frame difference value. For each. Subsequently, the above-described noise reduction apparatus determines the majority of the comparison results obtained by comparing the number of moving pixels and the number of still pixels with different threshold values to determine whether it is motion or noise, and obtains from the determination result. Noise removal is performed by setting a coefficient for noise in accordance with a signal indicating the degree of movement.

JP 2007-274067 A

  However, in the noise reduction apparatus using the majority decision based on the peripheral pixels described in Patent Document 1, there is a problem that an afterimage is conspicuous in the outline of the moving body and the image becomes unsightly. The principle of the problem will be described with reference to FIG.

  FIG. 21A shows an imaging screen in which an object 1001 having a high brightness with respect to the background 1002 is captured by the imaging apparatus. It is assumed that the object 1001 has a substantially uniform luminance and the background 1002 has a substantially uniform luminance. Next, it is assumed that the object 1001 at the position of FIG. 21A starts to move to the right on the screen, and moves to the position of FIG. 21C after one frame through the screen state shown in FIG. . Here, FIG. 21B shows the position of the object 1001 at the moment x seconds before the state of FIG. 21C (x is shorter than one frame interval), and as can be seen from the dotted line, FIG. It is located on the left side of the screen from 21 (c).

  Assume that the shutter speed of the imaging apparatus is x seconds. In the screen shown in FIG. 21D, in the area # 1, the background 1002 changes from the object 1001 to the object 1001 in the x seconds of FIGS. 21B to 21C, and in the area # 2, the object 1001 changes from the background 1002. Change. On the other hand, the region # 3 is at substantially the same luminance of the object 1001 for x seconds. As described above, due to the difference in exposure to the pixels due to the relationship between the shutter speed and the moving speed of the object, the image pickup signal, that is, the input video signal of the noise reduction device, is divided into a region # 1 and a region as shown in FIG. In # 2, it is expressed as a luminance intermediate between the background 1002 and the object 1001.

  Accordingly, the luminance difference between the screen shown in FIG. 21A and the screen shown in FIG. 21D after one frame is smaller in the region # 1 and the region # 2 than in the region # 3. For this reason, in the conventional frame cyclic noise reduction device, the region # 3 is determined to be a moving image portion, and the region # 1 and the region # 2 are determined to be a still portion (or an intermediate portion between a still image and a moving image). As shown in (e), the image portion 1003 of the region # 1 and the region # 2 has an intermediate luminance between the images of the object 1001 and the background 1002 of the input video signal (FIG. 21D).

  Furthermore, if the object 1001 continues to move, after one frame, the image by the luminance signal, that is, the input video signal to the noise reduction device, becomes as shown in FIG. 21 (f) for the same reason as described above. Here, the noise reduction apparatus determines the region # 1 and the region # 2 from the inter-frame difference value between the video signal of the previous frame shown in FIG. 21 (e) and the video signal of the current frame shown in FIG. 21 (f). Is determined as a still part (or an intermediate part between still and moving images), a video signal in which a contoured afterimage 1004 appears is output as shown in FIG.

  The region (pixel) that once becomes the contour-like afterimage 1004 remains as a still image until it converges while approaching the background after the next frame, causing unsightly and unpleasant video.

  In FIG. 21, the description is made in monochrome, but the same applies to the color difference component when there is a color difference. In addition, although the region # 1 and the region # 2 are expressed with uniform luminance, when there are a plurality of pixels in the region # 1 and the region # 2, the gradation actually ranges from the background luminance to the object luminance.

  As described above, in the noise reduction device using the majority decision based on the peripheral pixels described in Patent Document 1, since the peripheral pixels are included in the contour portion of the moving body, an appropriate determination cannot be performed. has no effect.

  The present invention has been made in view of the above points, and when performing frame cyclic noise reduction in which the cyclic coefficient is increased or decreased according to the motion determination result of the target pixel to reduce noise of the input video signal, the subject is displayed on the screen. An object of the present invention is to provide a noise reduction device that can reduce afterimages appearing in the contour of a moving body.

  In order to achieve the above object, the noise reduction device according to the first aspect of the present invention includes a plurality of noise reduction target pixels in a predetermined direction and a direction opposite to the predetermined direction with respect to a target pixel of noise reduction target of an input video signal input from the imaging device. The inter-frame difference value calculating means for calculating the inter-frame difference value between two alternative pixels or alternative pixel areas that are separated by the number of pixels, and the larger of the inter-frame difference values between the two alternative pixels or alternative pixel areas A first cyclic coefficient having a larger value as the selected inter-frame difference value is larger, and a second cyclic coefficient having a larger value as the selected inter-frame difference value is smaller. Cyclic coefficient calculation means for calculating the first video signal, first multiplication means for multiplying the input video signal by the first cyclic coefficient to generate a first post-multiplication signal, and output one frame before the input video signal A second multiplying unit for multiplying the image signal by the second cyclic coefficient to generate a second post-multiplication signal; and an adding unit for adding the first and second post-multiplication signals to generate an output video signal. And delay means for delaying the output video signal by one frame and generating an output video signal one frame before.

  In order to achieve the above object, the noise reduction apparatus according to the second aspect of the present invention provides an absolute value of an inter-frame difference value for each pixel of the input video signal and the input video signal one frame before and a motion determination threshold value. Based on the comparison result, a process for determining a moving pixel in the input video signal is performed for each frame of the input video signal, and imaging is performed based on the position of the barycentric pixel of each pixel determined to be a moving pixel in each frame. And a moving body speed calculating unit that calculates a moving speed of the moving body that is the subject of the element. The inter-frame difference value calculating unit calculates the number of pixels between the target pixel and the alternative pixel or the alternative pixel region. As the moving speed of the moving body calculated by the speed calculating means is faster, the moving body is more variable.

  Furthermore, in order to achieve the above object, a noise reduction device according to a third aspect of the present invention is directed to an imaging unit having a zoom function and a predetermined direction with respect to a target pixel for noise reduction of an input video signal input from the imaging unit. An inter-frame difference value calculating means for calculating inter-frame difference values of two alternative pixels or alternative pixel areas at positions separated from each other by a plurality of pixels in a direction opposite to the predetermined direction, and an input video signal and one frame A process for determining a motion pixel in the input video signal is performed for each frame of the input video signal based on the comparison result between the absolute value of the inter-frame difference value for each pixel of the previous input video signal and the threshold for motion determination. A moving body speed calculating means for calculating a moving speed of a moving body that is a subject of the image sensor based on a difference in position of the center-of-gravity pixel of each pixel determined as a moving pixel in each frame; Based on the information on either the shutter speed value or the zoom value obtained from the above, the information on both the shutter speed value and the zoom value, or the information on both the shutter speed value and the moving speed of the moving object, The number of pixels variable means for changing the number of pixels between the substitute pixel or the substitute pixel area and the larger inter-frame difference value of the difference values between the frames of the two substitute pixels or substitute pixel areas are selected and selected. A cyclic coefficient calculating means for calculating a first cyclic coefficient having a larger value as the inter-frame difference value is larger, and a second cyclic coefficient having a larger value as the selected inter-frame difference value is smaller; and an input video signal; A first multiplication means for multiplying the first cyclic coefficient to generate a first post-multiplication signal; a second multiplication coefficient multiplied by the output video signal one frame before the input video signal and the second cyclic coefficient; of A second multiplication means for generating a post-computation signal; an addition means for adding the first and second post-multiplication signals to generate an output video signal; and an output of the previous frame by delaying the output video signal by one frame. And delay means for generating a video signal.

  According to the present invention, frame-circulation-type noise is obtained by performing motion determination of a pixel of interest based on an inter-frame difference value of a substitute pixel or a substitute pixel region located at a position away from the pixel of interest by a plurality of pixels. When performing reduction, it is possible to reduce an afterimage that appears on the screen in the contour of a moving body that is a subject.

It is a block diagram of a 1st embodiment of a noise reduction device of the present invention. It is a figure which shows an example of the pixel number between the attention pixel and substitution pixel which are used for a motion determination in the motion determination part in FIG. It is a figure which shows an example of the difference value between frames in a motion determination part in FIG. 1, and a cyclic coefficient characteristic. It is a top view which shows an example of an imaging | photography condition. It is a figure which shows an example of the pixel count between the attention pixel and substitution pixel area | region used for a motion determination in the motion determination part in FIG. It is a figure which shows the other example of the difference value between frames in the motion determination part in FIG. 1, and the cyclic coefficient characteristic. It is a flowchart for operation | movement description of an example of the motion determination part in 1st Embodiment of the noise reduction apparatus of this invention. It is a flowchart for operation | movement description of the other example of the motion determination part in 1st Embodiment of the noise reduction apparatus of this invention. It is a block diagram of 2nd Embodiment of the noise reduction apparatus of this invention. It is a block diagram of carrying of the 3rd implementation of the noise reduction apparatus of this invention. It is a figure which shows the pixel structure of 1 frame. It is a flowchart for operation | movement description of the motion determination part in 3rd Embodiment of the noise reduction apparatus of this invention. It is a flowchart for operation | movement description of the motion determination part in the modification of 3rd Embodiment of the noise reduction apparatus of this invention. It is a figure which shows an example of the block division | segmentation of the screen of 1 frame. It is a figure which shows an example of the movement log | history of two moving bodies. FIG. 16 is a diagram (part 1) supplementing the explanation of the operation shown in FIG. 13 by taking the two moving bodies in FIG. 15 as an example; FIG. 16 is a diagram (part 2) supplementing the explanation of the operation shown in FIG. 13 by taking the two moving bodies of FIG. 15 as an example; It is a figure which shows the pixel structure of one block B (1, 4) in FIG. It is a figure which shows the pixel structural example in the block division different from FIG. It is a figure explaining patent document 1. It is problem explanatory drawing of the conventional frame cyclic noise reduction apparatus.

  Next, embodiments of the present invention will be described with reference to the drawings.

(First embodiment)
FIG. 1 shows a block diagram of a first embodiment of a noise reduction apparatus according to the present invention. In the figure, a noise reduction apparatus 100 according to the present embodiment includes an input terminal 101 to which a video signal including random noise obtained by imaging with an imaging element (not shown) is input, a subtraction unit 102, Frame memory 103, motion determination unit 104, multiplication units 105 and 106, addition unit 107, and video signal output terminal 108.

  The subtracting unit 102 takes a difference in luminance or color difference component between the video signal input via the input terminal 101 and the output video signal of the previous frame stored in the frame memory 103, and uses the difference signal as the motion determination unit 104. To send. The motion determination unit 104 determines a motion amount by a method described later based on the difference signal input from the subtraction unit 102, and generates a cyclic coefficient K (0 to 1) based on the motion amount.

  The multiplication unit 105 multiplies the video signal input via the input terminal 101 by the first cyclic coefficient (1-K) generated by the motion determination unit 104. The multiplying unit 106 multiplies the video signal of the previous frame by the second cyclic coefficient K generated by the motion determining unit 104. The adder 107 adds the post-multiplication signals output from the multipliers 105 and 106, outputs the output video signal to the video signal output terminal 108, supplies it to the frame memory 103, and stores one frame. The frame memory 103 supplies the output video signal to the video signal output terminal 108 to the multiplier 106 with a delay of one frame period.

  The noise reduction apparatus 100 according to the present embodiment is characterized in a method for determining a cyclic coefficient by motion determination by the motion determination unit 104, and the feature will be described below.

  First, a motion determination method in the motion determination unit 104 will be described with reference to FIGS.

  When the motion determination unit 104 performs the motion determination on the target pixel 150 illustrated in FIG. 2, the motion determination unit 104 does not determine the motion of the target pixel 150 itself, but instead of the alternative pixel L that is (p + 1) pixels away from the target pixel 150 in the left direction. Alternatively, the motion determination is performed based on the inter-frame difference value in one of the alternative pixels R of the alternative pixel R separated by (p + 1) pixels in the right direction. Here, of the left and right alternative pixels L and R, the alternative pixel having the larger inter-frame difference value is selected as a pixel for motion determination. This is because an alternative pixel having a larger inter-frame difference value can more reliably determine the movement of the subject. It is assumed that the number of pixels p is a value set as described later.

  FIG. 3 shows an example of inter-frame difference value versus cyclic coefficient characteristics in the motion determination unit 104. The motion determination unit 104 determines the cyclic coefficient K of the target pixel 150 according to the inter-frame difference value versus cyclic coefficient characteristic shown in FIG. 3 based on the inter-frame difference value of the selected alternative pixel. Accordingly, the motion determination unit 104 determines that the motion of the target pixel is larger as the inter-frame difference value of the selected alternative pixel is larger, and the value of the first cyclic coefficient (1-K) supplied to the multiplication unit 105 Is made large, and the value of the second cyclic coefficient K supplied to the multiplier 106 is made small. On the other hand, the motion determination unit 104 determines that the motion of the target pixel is smaller as the inter-frame difference value of the selected alternative pixel is smaller, and determines the value of the first cyclic coefficient (1-K) to be supplied to the multiplication unit 105. The value of the second cyclic coefficient K supplied to the multiplication unit 106 is set to be small and large.

  As a result, the noise reduction apparatus 100 according to the present embodiment applies the multiplied signal of the current frame obtained by multiplying the video signal from the image sensor input via the input terminal 101 by the first cyclic coefficient (1-K). On the other hand, the motion determination result by the motion determination unit 104 indicates the ratio of the post-multiplication signal of the previous frame obtained by multiplying the video signal of the previous frame from the multiplication unit 106 added by the addition unit 107 by the second cyclic coefficient K. It is possible to reduce the afterimage by setting the ratio to be smaller as the pixel movement is larger (decrease the amount of circulation), and to reduce the noise by increasing the ratio (increase the amount of circulation) as the movement of the pixel of interest is smaller. .

  As described above, in the conventional noise reduction device, the contour portion of the video signal of the moving object photographed as the subject has intermediate brightness with the background, so that the amount of motion cannot be determined correctly. On the other hand, the motion determination unit 104 according to the present embodiment uses, as the motion determination target pixel, a pixel that appears on either the left or right side of the target pixel and has a larger inter-frame difference value that deviates from the intermediate luminance. Thus, it is possible to accurately determine the movement of the target pixel.

  Next, a method for setting the number of pixels p in FIG. 2 will be described. The number of pixels p is preset to an appropriate value according to the shooting situation. For example, when the image sensor that supplies a video signal to the input terminal 101 in FIG. 1 is an image sensor of a surveillance camera, the number of pixels p is set based on the following conditions.

  As shown in the top view of FIG. 4, the monitoring camera 1101 captures an image of a subject 5 meters away from the monitoring camera 1101 with a horizontal pixel count of 1280 pixels and a shutter speed of 1/60 seconds in a shooting range 1102 with a horizontal angle of view of 60 °. It shall be. Here, it is assumed that the subject is a moving body that moves in the direction indicated by the arrow 1103 in FIG. 4 at an assumed moving speed of 2 meters / second (approximately steep walking), for example.

  The imaging range 1102 of the monitoring camera 1101 shown in FIG. 4 is an actual distance of about 5.77 (= 2 × 5 × tan 30 °) meters. The number of horizontal pixels of the monitoring camera 1101 is 1280 pixels. Accordingly, since 2 meters in the shooting range 1102 corresponds to about 443 (= 2 × 1280 ÷ 5.77) pixels, the moving speed (hereinafter referred to as moving speed) when the moving body that is the subject moves in the shooting range 1102 in the moving direction 1103. , Which is also referred to as a moving body speed) is 443 pixels / second.

  Now, since the shutter speed is 1/60 second, the region of intermediate luminance is about 7 pixels (= 443 pixels / 60). Between the alternative pixel L and the alternative pixel R shown in FIG. 2, there are (2p + 1) pixels. Therefore, the number of pixels p is set to “3”.

  Note that the alternative pixel is not limited to one pixel as shown in FIG. 2, and as shown in FIG. 5, n pixels (n is 2 or more) at a position away from the target pixel 150 by (p + 1) pixels to the left. It may be n pixels arranged in the horizontal direction, such as an alternative pixel area 151 made up of (natural number) or an alternative pixel area 152 made up of n pixels located at a distance of (p + 1) pixels in the right direction.

  Next, a method of determining the cyclic coefficient when the alternative pixel area is a plurality of pixels as shown in FIG. 5 will be described. FIG. 6 shows an example of inter-frame difference value versus cyclic coefficient characteristics in the motion determining unit 104 in this case. FIG. 7 is a flowchart for explaining an example of a cyclic coefficient determination method by the motion determination unit 104 in this case.

  In the alternative pixel area 151 or 152 in FIG. 5, the motion determination unit 104 in FIG. 1 performs an inter-frame difference at a preset m pixel (m is a natural number less than n) or more among n pixels constituting the alternative pixel area. It is determined whether or not the value is greater than or equal to the maximum threshold value TH1 (step S1). When there are m or more pixels having an inter-frame difference value equal to or greater than the threshold TH1, the motion determination unit 104 determines that the target pixel 150 is a pixel having a large motion and sets the cyclic coefficient K to “0” (step S2). .

  If there are no m pixels or more of the n pixels whose inter-frame difference value is greater than or equal to the threshold value TH1, in the alternative pixel area 151 or 152, the frame is set in m pixels or more that are set in advance among the n pixels constituting the alternative pixel area. It is determined whether or not the difference value is greater than or equal to a threshold value TH2 (TH2 <TH1) (step S3). When there are m or more pixels whose inter-frame difference value is equal to or greater than the threshold TH2, the motion determination unit 104 determines that the target pixel 150 is a pixel having a relatively large motion, and the cyclic coefficient K is a small value illustrated in FIG. K1 is set (step S4).

  Subsequently, in the alternative pixel area 151 or 152, the threshold THj (THj <... <TH2 <TH1) in which the inter-frame difference value of m pixels or more set in advance among the n pixels constituting the alternative pixel area is the minimum. It is determined whether or not the above is true (step S5). When there are m or more pixels whose inter-frame difference value is equal to or greater than the threshold THj, the motion determination unit 104 determines that the target pixel 150 is a pixel having a relatively small motion, and the cyclic coefficient K is a large value illustrated in FIG. Ki is set (step S6). On the other hand, if it is determined in step S5 that the number of pixels having an inter-frame difference value equal to or greater than the threshold THj is not greater than m, the motion determination unit 104 determines that the pixel of interest 150 is a pixel with little motion and sets the cyclic coefficient K. The maximum value Kj shown in FIG. 6 is set (step S7).

  In this way, since the motion determination unit 104 uses the majority determination result of a plurality of pixels for motion determination, even when random noise appears in the alternative pixel areas 151 and 152, the influence of random noise is eliminated. And the magnitude of movement can be determined with high accuracy.

  Note that the method of taking the alternative pixel and the alternative pixel region is not limited to that illustrated in FIGS. 2 and 5. For example, pixels on a different line from the target pixel may be included in the alternative pixel or the alternative pixel region.

  By the way, in the method described above, when there is a moving body having a spatial size of (2p + 1) pixels or less, there is a possibility that it is determined as a stationary part or a simple stationary part is determined as a motion.

  Therefore, in order to eliminate such a possibility, in the modification of the present embodiment, the motion determination unit 104 does not perform determination based only on the alternative pixel or the alternative pixel region, but, for example, according to the procedure illustrated in the flowchart of FIG. Judged together with the result of.

  In FIG. 8, the motion determination unit 104 determines whether or not the inter-frame difference value at the target pixel 150 is greater than or equal to THx (step S11). The threshold value THx is a sufficiently large value that can be determined as a moving part only by the result of the inter-frame difference value of the target pixel 150. When the motion determination unit 104 determines that the inter-frame difference value at the target pixel 150 is greater than or equal to the maximum threshold value THx, the motion determination unit 104 determines that the target pixel 150 is a motion part and sets the cyclic coefficient K to “0”. (Step S12).

  On the other hand, when the motion determination unit 104 determines in step S11 that the interframe difference value at the target pixel 150 is not equal to or greater than the maximum threshold value THx, the interframe difference value at the target pixel 150 is equal to or less than the minimum threshold value THy. It is determined whether or not (step S12). The threshold value THy is a sufficiently small value that can be determined as a still part only by the result of the inter-frame difference value of the target pixel 150.

  When the motion determination unit 104 determines that the inter-frame difference value at the target pixel 150 is equal to or smaller than the minimum threshold value THy, the motion determination unit 104 determines that the target pixel 150 is a stationary part and sets the cyclic coefficient K close to “1”. A value Kj (for example, 0.8) is set (step S14). When the motion determination unit 104 determines that the inter-frame difference value at the target pixel 150 is not less than or equal to the minimum threshold value THy, the target pixel 150 determines whether the target pixel is a still part or a motion part only by the inter-frame difference value of the target pixel. Since this is not possible, the cyclic coefficient is determined based on the substitute pixel or the substitute pixel region (step S15). In step S15, when the alternative pixel shown in FIG. 2 is used, the cyclic coefficient is determined according to the inter-frame difference value versus the cyclic coefficient characteristic shown in FIG. 3, and when the alternative pixel areas 151 and 152 shown in FIG. The cyclic coefficient is determined according to the flowchart shown in FIG. 7 and the inter-frame difference value versus cyclic coefficient characteristic shown in FIG.

  In the flowchart of FIG. 8, the determination by the pixel of interest may be the determination at (2p + 1) pixels around the pixel of interest. However, in this case, it is necessary to replace steps S11 and S13 shown in FIG. 8 with steps for performing the following processing. That is, step S11 is a step of determining whether or not the inter-frame difference value equal to or greater than a preset s pixel (s is a natural number equal to or less than 2p + 1) among (2p + 1) pixels in the horizontal direction of the target pixel is equal to or greater than the threshold value THx. Replace with Further, step S13 shown in FIG. 8 is replaced with a step of determining whether or not the inter-frame difference value equal to or larger than the preset s pixel among (2p + 1) pixels around the target pixel is equal to or smaller than the threshold value THy.

  As described above, according to the present embodiment, the amount of motion of the target pixel 150 is determined based on the inter-frame difference value of the alternative pixel or the alternative pixel region that is horizontally separated from the target pixel 150 by the number of pixels p. Therefore, it is possible to reduce the afterimage of the moving body that appears in a contour shape on the imaging screen.

(Second Embodiment)
Next, a second embodiment of the present invention will be described. FIG. 9 shows a block diagram of a second embodiment of the noise reduction apparatus according to the present invention. In the figure, the same components as those in FIG.

  As shown in FIG. 9, the noise reduction apparatus 200 according to the present embodiment includes an input terminal 101 to which a video signal including random noise obtained by imaging with an imaging element (not shown) is input, and subtraction. A unit 102, a frame memory 103, multiplication units 105 and 106, an addition unit 107, a video signal output terminal 108, an imaging setting value acquisition unit 201, and a motion determination unit 204 are configured.

  The imaging setting value acquisition unit 201 includes a shutter speed acquisition unit 202 that acquires a shutter speed of an imaging unit (not shown), and a zoom value acquisition unit 203 that acquires a zoom value of the imaging unit. The motion determination unit 204 adaptively variably sets the number of pixels p described above according to not only the inter-frame difference value from the subtraction unit 102 but also the shutter speed and zoom value acquired by the imaging setting value acquisition unit 201. To calculate the cyclic coefficient K.

  In the noise reduction apparatus 100 of the first embodiment described above, since the number of pixels p is a fixed value, the number of pixels p is too small depending on the moving body speed, and the alternative pixel (or alternative pixel region) is an intermediate luminance region. And the amount of motion cannot be determined correctly, or on the other hand, the number of pixels p is too large and the alternative pixel (or alternative pixel region) may be removed from the moving object. Therefore, the noise reduction apparatus 200 according to the present embodiment is characterized in that the number of pixels p is adaptively changed in order to improve the determination accuracy of the motion amount.

  As described above, an afterimage that is an intermediate luminance between the moving object that is the subject and the background on the screen is caused by a difference in exposure to pixels due to the relationship between the shutter speed and the moving speed of the moving object. Therefore, if the moving speed of the moving body (the number of pixels that advance on the screen per unit time) is the same, the intermediate luminance region becomes narrower as the shutter speed increases. Therefore, the motion determination unit 204 in the present embodiment controls to decrease the number of pixels p as the shutter speed increases.

  In addition, even with a moving body that is actually moving at the same speed, the larger the zoom value, the greater the number of pixels that advance per unit time. Therefore, the motion determination unit 204 in the present embodiment controls to increase the number of pixels p as the zoom value increases.

  For example, under the conditions of the first embodiment, the distance between the subject and the camera is 5 meters, the number of horizontal pixels of the camera is 1280 pixels, the assumed moving body speed is 2 meters / second (approximately steep walking), and the shutter speed is x seconds. If the zoom value changes to y times the previous condition, the motion determination unit 204 sets the number of pixels p based on the following equation.

p = 443 · x · y / 2 (rounded down) (1)
Here, “443” in the above equation is “443”, which is the number of pixels per second that the moving body moves. Thereby, for example, when the shutter speed x changes to 1/100 second and the zoom value y changes three times, the number of pixels p is set to “6” from the above equation.

  As described above, according to the present embodiment, the number of pixels p calculated by the motion determination unit 204 based on the inter-frame difference value is reduced as the shutter speed acquired by the shutter speed acquisition unit 202 increases. The number of pixels p can be set to a more appropriate value by variably adjusting to a value and variably adjusting the number of pixels p to a larger value as the zoom value acquired by the zoom value acquisition unit 203 is larger. The cyclic coefficient can be more appropriately determined according to the body speed and the zoom value, and the afterimage can be reduced with high accuracy.

(Third embodiment)
Next, a third embodiment of the present invention will be described. FIG. 10 shows a block diagram of a third embodiment of the noise reduction apparatus according to the present invention. In the figure, the same components as those in FIG.

  As shown in FIG. 10, the noise reduction apparatus 300 according to the present embodiment includes an input terminal 101 to which a video signal including random noise obtained by imaging with an imaging element (not shown) is input, and subtraction. Unit 102, frame memory 103, multiplication units 105 and 106, addition unit 107, video signal output terminal 108, shutter speed acquisition means 301, contour pixel number calculation unit 302, barycentric pixel recording unit 303, , And a motion determination unit 304.

  The shutter speed acquisition unit 301 acquires the shutter speed of the imaging unit (not shown). As will be described later, the contour pixel number calculation unit 302 obtains the horizontal center-of-gravity pixel of the motion pixel in one frame from the inter-frame difference signal supplied from the subtraction unit 102, and then moves the moving body speed based on the center-of-gravity pixel. And the number of pixels p is obtained from the moving body speed. The contour pixel number calculation unit 302 supplies the calculated pixel number p to the motion determination unit 304, records the calculated centroid pixel in the centroid pixel recording unit 303, and is recorded from the centroid pixel recording unit 303 when calculating the moving body velocity. The center-of-gravity pixel is read.

  Based on the inter-frame difference signal supplied from the subtraction unit 102 and the pixel number p supplied from the contour pixel number calculation unit 302, the motion determination unit 304 sets the cyclic coefficient K in the same manner as the motion detection unit 104 in FIG. The cyclic coefficient K is calculated by the above operation, the cyclic coefficient K is output to the multiplier 106, and the cyclic coefficient (1-K) is output to the multiplier 105.

  Next, a method for calculating the speed of the moving object and the number of pixels p in the contour pixel number calculation unit 302 will be described with reference to the flowcharts of FIGS. FIG. 11 shows all pixels of one frame (horizontal n pixels, vertical m pixels) of the input video signal. FIG. 12 is a flowchart illustrating an example of a procedure for obtaining the number of pixels p.

  The contour pixel number calculation unit 302 first has a vertical direction i-th (i is a natural number of m or less) and horizontal j-th (j is a natural number of n or less) pixel X (j is a natural number of n or less) of one frame of the input video signal shown in FIG. i, j) and the absolute value | X (i, j) −Y (i, j) | of the difference between the pixel Y (i, j) (not shown) at the same position in the previous frame is set in advance. It is determined whether or not the threshold value TH is greater. If it is larger, the pixel is determined as a pixel that has moved (hereinafter referred to as a moving pixel), and if not, the pixel is determined as a still pixel (step S21). The contour pixel number calculation unit 302 repeats the process in step S21 until all pixels in one frame are completed (step S22).

  Next, the contour pixel number calculation unit 302 obtains a horizontal center-of-gravity pixel by the following equation for the motion pixels in one frame, and records the center-of-gravity pixel in the center-of-gravity pixel recording unit 303 (step S23).

Centroid pixel = (sum of horizontal coordinate values j of the moving pixels) / total number of moving pixels (2)
Subsequently, the contour pixel number calculation unit 302 reads out the centroid pixel of the motion pixel one frame before from the centroid pixel recording unit 303, and calculates the moving body speed (unit: pixels / second) by the following equation (step S24). .

Moving body speed = (centroid pixel of current frame-1 centroid pixel before frame) / frame interval
(3)
However, if there is no moving pixel in the current frame, the moving body speed is set to “0” because there is no intermediate luminance region. If there is no moving pixel only one frame before, the moving body speed is calculated by the following equation.

Moving object velocity = (n / 2) − | centroid pixel of current frame− (n / 2) | (4)
That is, the moving body speed is applied based on the moving distance from the left and right ends of one frame to the center-of-gravity pixel.

  Subsequently, the contour pixel number calculation unit 302 calculates the pixel number p based on the following equation based on the shutter speed obtained from the shutter speed acquisition unit 301 and the moving body speed calculated in step S24 (step S25).

p = (moving body speed × shutter speed) / 2 (rounded down) (5)
For example, when the moving body speed is 400 (pixels / second) and the shutter speed is 1/60 (seconds), the number of pixels p is 3 pixels from the equation (5).

  Finally, the contour pixel number calculation unit 302 sends the pixel number p to the motion determination unit 304 (step S26).

  The motion determination unit 304 refers to the pixel number p calculated by the contour pixel number calculation unit 302 with reference to the next frame to perform motion determination, and calculates the cyclic coefficient K. The operation of the motion determination unit 304 is the same as that of the motion determination unit 104 described with reference to FIGS.

  In the above description, the contour pixel number calculation unit 302 calculates the number of pixels p every frame and sends the latest value to the motion determination unit 304. For example, an average of several values such as three frames is calculated. It may be adopted.

  As described above, the moving body speed can be calculated to obtain the number of pixels p. However, when there are a plurality of moving objects in one screen, the moving object speed cannot be obtained correctly at the center of gravity of the moving pixels.

  FIG. 15 shows an example of a movement history for each frame of two moving bodies A and B traveling in different directions. As shown in FIG. 15, when there are two moving bodies A and B traveling in different directions, the method according to the flowchart shown in FIG. 12 cancels the moving body speed, so that the number of pixels p can be calculated correctly. Can not.

  Therefore, in the modification of the third embodiment described below with reference to FIGS. 13 to 18, the calculation amount increases, but the number of pixels p is accurately calculated even when there are a plurality of moving bodies.

  FIG. 13 is a flowchart illustrating an example of a procedure for obtaining the number of pixels p in the modification. The contour pixel number calculation unit 302 first determines the movement of each pixel in the block (step S31). For example, as shown in FIG. 14, the block in step S31 is a total of 16 rectangular areas B (1, 1) to B (4, which are obtained by dividing one frame of the input video signal into 4 blocks in the horizontal direction and 4 blocks in the vertical direction. 4).

  In step S31, the contour pixel number calculation unit 302 determines that the first block B (1, 1) of the 16 blocks is g-th in the vertical direction of one frame (g is equal to or less than m / 4). Natural number), the difference between the h-th pixel X (g, h) in the horizontal direction (h is a natural number equal to or less than n / 4) and the pixel Y (g, h) (not shown) at the same position one frame before. It is determined whether or not the absolute value | X (g, h) −Y (g, h) | is greater than a preset threshold value TH. If it is larger, the pixel is determined to be a moving pixel (hereinafter referred to as a moving pixel), and if not, the pixel is determined to be a still pixel. The contour pixel number calculation unit 302 repeats the process of step S31 described above until all pixels of one block are completed (step S32). That is, the pixels are input in the order of X (1,1), X (1,2),..., X (1, n), X (2,1), X (2,2),. In this case, the pixels up to the pixel X (m / 4, n / 4) are all the pixels of the first block B (1, 1).

  Next, the contour pixel number calculation unit 302 calculates the horizontal center-of-gravity pixel by the following equation for the motion pixels in one block, and records the calculated center-of-gravity pixel in the center-of-gravity pixel recording unit 303 (step S33).

Center-of-gravity pixel = (sum of horizontal coordinate values h of moving pixels in one block) / (in one block
Total number of motion pixels) (6)
Subsequently, the contour pixel number calculation unit 302 reads out the centroid pixel of the motion pixel in the corresponding block one frame before from the centroid pixel recording unit 303, and calculates the moving body speed (unit: pixels / second) by the following equation ( Step S34).

Moving object speed = | centroid pixel of current frame−1centroid pixel before frame | / frame interval
(7)
However, when there is no moving pixel in either the previous frame or the corresponding block of the current frame, the moving body speed is set to “0”.

  Subsequently, the contour pixel number calculation unit 302 calculates the pixel number p based on the following equation based on the shutter speed obtained from the shutter speed acquisition unit 301 and the moving body speed calculated in step S34 (step S35).

p = (moving body speed × shutter speed) / 2 (rounded down) (8)
Then, the contour pixel number calculation unit 302 sends the pixel number p to the motion determination unit 304 (step S36). Since the number of pixels p calculated in step S35 is the number of pixels for one block, the contour pixel number calculation unit 302 continues until the calculation of the number of pixels p for all the remaining blocks in one frame is completed. Steps S31 to S36 are repeated (Step S37). When the calculation of the number of pixels p is completed for all blocks of one frame, the processing is completed.

  Next, using the two moving bodies A and B shown in FIG. 15 as an example, the explanation of the operation will be supplemented using FIG. 16 and FIG. FIG. 16 and FIG. 17 show the area where the motion is detected in frame 2 and frame 3 and the barycentric position in the block, respectively.

  That is, in frame 2, the area 2a is a motion detection area based on the difference in which the moving object A newly appears from the background due to the movement of the moving object A during the period from the frame 1 to the frame 2, and the area 1a is moved. This is a motion detection region based on a difference in which the body A disappears. The white part between them passes only between the moving body A itself between frames, so that the difference is small and it does not become a motion detection region. The crosses indicate the in-block barycentric positions of the block B (2, 4) and the block B (3, 4).

  In frame 2, area 2b is a motion detection area based on a difference in which moving body B newly appears from the background due to movement of moving body B during the period from frame 1 to frame 2, and area 1b moves. This is a motion detection region based on the difference in which the body B disappears. The crosses indicate the in-block barycentric positions of the block B (1, 1) and the block B (1, 2).

  Similarly, in the frame 3, the area 3 a is a motion detection area based on a difference in which the moving object A newly appears from the background due to the movement of the moving object A during the period from the frame 2 to the frame 3. This is a motion detection area based on the difference in which the moving object A disappears. The white part between them passes only between the moving body A itself between frames, so that the difference is small and it does not become a motion detection region. The crosses indicate the in-block barycentric positions of the block B (2, 4) and the block B (3, 4).

  In frame 3, area 3b is a motion detection area based on a difference in which moving body B newly appears from the background due to movement of moving body B during the period from frame 2 to frame 3, and area 2b moves. This is a motion detection region based on the difference in which the body B disappears. The crosses indicate the in-block barycentric positions of the block B (1, 1) and the block B (1, 2).

  Thus, as can be seen from FIGS. 16 and 17, each of the blocks B (1,1), B (1,2), B (2,4), and B (3,4) has two mobile objects. The moving body speed and the number of pixels p can be obtained without interfering with the region where the motion is detected.

  The motion determination unit 304 uses the pixel number p for each block obtained as described above for motion determination of all the pixels in the block. For example, in block B (1, 4), there are n · m / 16 pixels X (1, (3n / 4) +1) to pixels X (m / 4, n) shown in FIG. The moving body speed in the block B (1, 4) is obtained after the last pixel X (m / 4, n) in the block is input.

  Therefore, in order to perform the motion determination from the first pixel X (1, (3n / 4) +1), the motion determination unit 304 can store the m / 4 line data in the memory, and the moving body speed can be increased. Motion determination can be performed after it is obtained. Further, in the case of a configuration without a memory, the obtained number of pixels p may be adopted in the next frame. The operation of the motion determination unit 304 is the same as the operation of the motion determination unit 104 described with reference to FIGS.

  In the above description, the divided blocks are fixed, but may vary for each pixel for which motion is determined. For example, among the pixels in the block B (1, 4) shown in FIG. 18, the pixels X (m / 4, (3n /) adjacent to the block B (1, 3) and the block B (2, 4), respectively. 4) When performing motion determination for +1), as shown in FIG. 19, horizontal n / 4 pixels and vertical m / 4 pixels centered on the pixel X (m / 4, (3n / 4) +1) One block may be configured, and the number of pixels p obtained in the block may be adopted.

  The present invention is not limited to the above-described embodiments and modifications. For example, in the second and third embodiments shown in FIGS. Although it has been described that the number of pixels p in the horizontal direction between the substitute pixel or the substitute pixel region is calculated, the shutter speed is fixed to a standard value in advance, and the number of pixels p is calculated only by the moving body speed. May be. In the second embodiment shown in FIG. 9, the number of pixels p may be changed according to only the zoom value or only the shutter speed. Further, in the third embodiment shown in FIG. 10, the number of pixels p is determined according to the shutter speed and the moving body speed by the equation (5). However, in the present invention, the number of pixels p is converted into the zoom value and the moving body. The method of changing according to speed is also included. Further, the cyclic coefficient K may be a characteristic opposite to the characteristic shown in FIG. 3 or FIG. 6 with respect to the inter-frame difference value. However, in that case, the motion determination units 104, 204, and 304 need to supply the first cyclic coefficient K to the multiplication unit 105 and supply the second cyclic coefficient (1−K) to the multiplication unit 106. is there.

  In addition, although each said Example mainly demonstrated the example of the noise reduction in a horizontal direction, in order to simplify description, of course, this invention is applicable also to a perpendicular direction or a diagonal direction.

  Further, the present invention includes a program for causing a computer to realize the functions of the above-described apparatus. These programs may be read from a recording medium and loaded into a computer, or may be transmitted via a communication network and loaded into a computer.

100, 200, 300 Noise reduction device 101 Video signal input terminal 102 Subtraction unit 103 Frame memory 104, 204, 304 Motion determination unit 105, 106 Multiplication unit 107 Addition unit 108 Video signal output terminal 150 Pixels of interest 151, 152 Alternative pixel region 201 Imaging setting value acquisition units 202 and 301 Shutter speed acquisition unit 203 Zoom value acquisition unit 302 Outline pixel number calculation unit 303 Center of gravity pixel recording unit 1101 Surveillance camera

Claims (3)

Two alternative pixels or alternative pixel regions that are located at a distance of a plurality of pixels in a predetermined direction and a direction opposite to the predetermined direction with respect to a target pixel for noise reduction of an input video signal input from an imaging device Inter-frame difference value calculating means for calculating the inter-frame difference value, respectively,
Selecting the larger inter-frame difference value of the two alternative pixels or the inter-pixel difference value of the alternative pixel region, and the larger the selected inter-frame difference value, the larger the first cyclic coefficient, Cyclic coefficient calculation means for calculating a second cyclic coefficient having a larger value as the selected inter-frame difference value is smaller;
First multiplication means for multiplying the input video signal by the first cyclic coefficient to generate a first post-multiplication signal;
Second multiplying means for generating a second post-multiplication signal by multiplying the output video signal one frame before the input video signal and the second cyclic coefficient;
Adding means for adding the first and second multiplied signals to generate an output video signal;
And a delay means for delaying the output video signal by one frame and generating the output video signal of the previous frame. Processing for determining a motion pixel in the input video signal based on a comparison result between an absolute value of an inter-frame difference value for each pixel of the input video signal and the input video signal one frame before and a threshold value for motion determination; A moving body that is performed for each frame of the input video signal and calculates a moving speed of a moving body that is a subject of the image sensor based on a difference in position of the center-of-gravity pixel of each pixel determined to be a moving pixel in each frame It further has a speed calculation means,
The inter-frame difference value calculating means increases the number of pixels between the target pixel and the substitute pixel or the substitute pixel area as the moving speed of the moving body calculated by the moving body speed calculating means increases. The noise reduction device according to claim 1, wherein the noise reduction device is variable. An imaging means having a zoom function;
Two alternative pixels or alternative pixels that are located at a distance of a plurality of pixels in a predetermined direction and in a direction opposite to the predetermined direction with respect to the target pixel for noise reduction of the input video signal input from the imaging means An inter-frame difference value calculating means for calculating an inter-frame difference value of the region;
Processing for determining a motion pixel in the input video signal based on a comparison result between an absolute value of an inter-frame difference value for each pixel of the input video signal and the input video signal one frame before and a threshold value for motion determination; A moving body that is performed for each frame of the input video signal and calculates a moving speed of a moving body that is a subject of the image sensor based on a difference in position of the center-of-gravity pixel of each pixel determined to be a moving pixel in each frame Speed calculation means;
Information on one of the shutter speed value and zoom value obtained from the imaging means, information on both the shutter speed value and the zoom value, or information on both the shutter speed value and the moving speed of the moving body Based on the pixel number variable means for changing the number of pixels between the pixel of interest and the replacement pixel or the replacement pixel region,
Selecting the larger inter-frame difference value of the two alternative pixels or the inter-pixel difference value of the alternative pixel region, and the larger the selected inter-frame difference value, the larger the first cyclic coefficient, A cyclic coefficient calculation means for calculating a second cyclic coefficient having a larger value as the selected inter-frame difference value is smaller;
First multiplication means for multiplying the input video signal by the first cyclic coefficient to generate a first post-multiplication signal;
Second multiplying means for generating a second post-multiplication signal by multiplying the output video signal one frame before the input video signal and the second cyclic coefficient;
Adding means for adding the first and second multiplied signals to generate an output video signal;
And a delay means for delaying the output video signal by one frame and generating the output video signal of the previous frame.