JP2006237716A - Image processor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processor capable of executing high-quality 3D-DNR processing by adaptively executing the proper 3D-DNR only to a part requiring the 3D-DNR in one image. <P>SOLUTION: A data selector 1 stores respective image data for three frames of a continuous input image by sorting them into FRBs 2, 3, and 4. A comparator 5 calculates a value of an autocorrelation coefficient for each pixel of the same coordinate on the basis of the image data for three frames stored in the FRBs 2, 3, and 4. The comparator 5 compares the values of autocorrelation coefficients between respective frames and determines the correlation between the respective frames. Movement determination is executed by classifying the results of the correlation determination among three frames into four patterns. A 3D-DNR unit 6 is informed of the pixel in which the movement determination result is judged to be stillness or movement stop. The 3D-DNR unit 6 executes the 3D-DNR processing to the informed pixel or an assembly of the informed pixels. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an image processing apparatus suitably used for an apparatus or the like that needs to write a large amount of image data in a storage means such as a memory cell. More specifically, the present invention adaptively performs three-dimensional noise reduction processing on image data. The present invention relates to an image processing apparatus for improving image quality.

  As a current technique for reducing noise in image data, as disclosed in Patent Documents 1 and 2 below, it is widely known to perform three-dimensional noise reduction processing on image data.

  Patent Document 1 discloses a three-dimensional noise reduction processing technique for the purpose of providing a video signal processing circuit and a video processing method that can improve image quality with a simple configuration. That is, the motion detection circuit 102 detects the presence / absence of video motion based on the composite video signal VD and the video signal of two frames of the video signal one frame before from the frame memory 101. The three-dimensional Y / C separation circuit 103 separates the luminance signal Y1 and the color signal C1 from the composite video signal VD based on the motion signal MS1 output from the motion detection circuit 102. The motion detection circuit 107 detects the presence / absence of video motion based on the luminance signal Y1 output from the three-dimensional Y / C separation circuit and the luminance signal of two frames of the luminance signal one frame before from the frame memory 105. . The synthesis circuit 108 synthesizes the motion signals MS1 and MS2 output from the motion detection circuit 102 and the motion detection circuit 107. The three-dimensional noise reduction circuit 106 reduces the noise of the luminance signal Y1 output from the three-dimensional Y / C separation circuit 103 based on the output signal of the synthesis circuit 108.

In Japanese Patent Laid-Open No. 2004-260260, video noise including a three-dimensional noise reduction process that aims to prevent degradation of still image quality due to noise cancellation and to reduce video noise caused by noise reduction on the time axis. Reduction techniques are disclosed. That is, in the video signal processing circuit, when the input video signal is a still image, the motion detection circuit 3 turns off the operation of the horizontal noise canceller (horizontal NC) 3 or applies noise cancellation (NC) to the input video signal. An NC control signal is supplied to the horizontal NC 3 so as to be weakened, while a frame / field noise reduction (NR) circuit (noise reduction is performed using the correlation of patterns between frames or fields) 1 is noise reduction (NR) The NR control signal is supplied to the NR circuit 1 so as to be applied to the input video signal. If the input video signal is a moving image, the NR control signal is supplied to the NR circuit 1 so that the operation of the NR circuit 1 is turned off or compared with a still image to weaken the NR applied to the input video signal. An NC control signal is supplied to the horizontal NC 3 so as to strengthen the application of NC to the input video signal (moving image). Here, whether the image is a still image or a moving image is detected by the motion detection circuit 5 from the difference between the video signal and the video signal delayed by one frame or one field (one frame or one field previous video signal). That's it.
JP 2000-032494 A JP 11-0669202 A

  However, the conventional technique has the following problems. That is, when 3D noise reduction processing (hereinafter referred to as 3D-DNR) is performed on an image, the 3D-DNR processing is conventionally performed on the entire image in units of frames or fields. 3D-DNR is uniformly applied to the non-existing portion, and in the image after the 3D-DNR processing, there is a problem that fine details are lost in a portion that originally has no noise. Further, the conventional 3D-DNR process has a problem that fixed noise such as beat noise cannot be removed. Further, in the conventional motion determination, the difference between the image delayed by one field or one frame and the current image is simply taken and compared with the threshold value, and the delicate correlation between the previous and subsequent images is completely taken into consideration. Therefore, there is a problem in that 3D-DNR processing cannot be performed by capturing an image that moves delicately, such as a movement toward a stop. Furthermore, there is a problem that an afterimage is generated in the processed image because an appropriate 3D-DNR cannot be applied to a fast moving image.

  The present invention has been made in view of the above-mentioned circumstances, and adaptively performs an appropriate 3D-DNR only on a necessary portion of the 3D-DNR in one image. An object of the present invention is to provide an image processing apparatus that performs high-quality 3D-DNR processing.

  In order to solve the above-mentioned problem, the invention according to claim 1 relates to an image processing device, and is an image processing device that performs three-dimensional noise reduction processing on image data. Image selection means for distributing input image data to storage means for storing; calculation means for calculating an autocorrelation coefficient value for each pixel of the image data between the frames; and for each pixel based on the calculated autocorrelation coefficient value Determining means for determining whether or not to perform the three-dimensional noise reduction process; and image processing means for adaptively performing the three-dimensional noise reduction process on a pixel or a set of pixels determined to perform the three-dimensional noise reduction process. It is characterized by having.

  The invention according to claim 2 relates to the image processing apparatus according to claim 1, wherein the determining means determines the motion of the image from the calculated autocorrelation coefficient value, and the stationary motion is determined from the motion determination result of the image. Alternatively, it is characterized in that a pixel determined to stop moving is determined as a pixel on which a three-dimensional noise reduction process is performed.

  The invention according to claim 3 relates to the image processing apparatus according to claim 1, wherein the determination unit calculates an autocorrelation coefficient value between a target pixel of an arbitrary frame and a peripheral pixel of another frame. It is characterized in that a motion is determined by reflecting it in the autocorrelation coefficient value between the pixels, and a pixel determined to be still or stopped from the result of the motion determination is determined as a pixel for performing a three-dimensional noise reduction process. .

  Furthermore, the invention according to claim 4 relates to the image processing apparatus according to claim 2 or 3, wherein the determination means determines that autocorrelation coefficient values between an arbitrary frame and the two frames before and after this are correlated. If it is determined, the target pixel is determined to be stationary, the autocorrelation coefficient value between the arbitrary frame and the previous frame is determined to be uncorrelated, and the arbitrary frame and this If the autocorrelation coefficient value between the subsequent frames is determined to be correlated, it is determined that the pixel of interest has stopped moving, and the pixels that have been determined to be stationary or stopped are subjected to three-dimensional noise reduction. It is characterized in that it is determined as a pixel to be processed.

  According to the configuration of the present invention as described above, as a first effect, when performing the 3D-DNR process, an autocorrelation function with the pixel values of the preceding and succeeding frames is calculated, and pixels with high correlation are obtained. Since the 3D-DNR process is selectively performed only on the part, the process is not performed at a place where the process is unnecessary, and the image quality of the processed image can be improved. Further, as a second effect, since 3D-DNR is performed only in a necessary portion, the processing speed can be increased. Further, as a third effect, since the motion determination is performed using the correlation between the images of the three frames, it is possible to perform the 3D-DNR process by capturing an image that moves slightly, such as a motion toward a stop. It is to become. Further, as a fourth effect, by reflecting the autocorrelation coefficient value between the target pixel and the surrounding pixels in the autocorrelation coefficient value between the target pixels, it becomes possible to distinguish between noise and an image and to remove fixed beat noise. Is to be able to.

An object of the present invention is to provide an image processing apparatus that performs high-quality 3D-DNR processing by adaptively performing appropriate 3D-DNR only for a place where 3D-DNR is necessary in one image. When performing 3D-DNR processing of an image, pixel data for three frames is stored, a correlation between each frame image is investigated for each pixel, and image motion is determined. Is determined to be 3D-DNR processing adaptively for each pixel or for each set of pixels. It is known that an image generally has a high correlation between pixels in the horizontal direction and the vertical direction. Therefore, in the processed image and the previous and next frame images, in addition to the correlation of the target pixel or the correlation of the target pixel, the correlation between the target pixel and the surrounding pixels is investigated and the motion of the image is determined. By performing 3D-DNR processing adaptively on a portion having high correlation between pixels determined to be still or stopped, 3D-DNR processing is performed only on the portion of the image that requires 3D-DNR processing. In addition to being able to implement, it is possible to improve the quality of the processed image, speed up the 3D-DNR process, and eliminate fixed beat noise.
Embodiments of the present invention will be described below with reference to the drawings. The description will be made specifically with reference to examples.

  First, a first embodiment of the present invention will be described. FIG. 1 is a block diagram showing an electrical configuration of the image processing apparatus according to this embodiment. The image processing apparatus of this embodiment is connected to an external system bus 7 or the like, and a data selector 1 that distributes input image data for three frames to a frame buffer described below, and the allocated image data for one frame. Frame buffers (hereinafter referred to as FRB) 2, 3, and 4 as storage means for holding each frame, image data for one frame held in each FRB is read for each pixel, and the autocorrelation coefficient value between frames is self Computation by a correlation function, image motion determination is performed based on the result, and based on the motion determination result, a comparator 5 that determines whether or not to perform 3D-DNR processing for each pixel on the input image, and determination by the comparator 5 As a result, a 3D-DNR unit that performs 3D-DNR processing for each pixel or a set of those pixels on the input image And it has a door. Here, each pixel data of the image data may constitute color image data having color information, or may constitute monochrome image data.

式（１）において、ｎはサンプリングポイント数、Ｎはデータ数、Ｙ（ｎ）、Ｙ（ｎ−ｋ）は画素データ、ｋはずらし時間順を表す。なお、自己相関係数値は絶対値をとるものとする。 In this image processing apparatus, when performing 3D-DNR processing, the data selector 1 first sorts and stores image data FR0, FR1, and FR2 for three consecutive input images into FRB2, FRB3, and FRB4. . As shown in FIG. 2, the comparator 5 calculates an autocorrelation coefficient value for each pixel of the same coordinate based on the image data for three frames stored in FRB2, FRB3, and FRB4. The autocorrelation coefficient value is generally calculated by an autocorrelation function represented by the equation (1).

In Expression (1), n represents the number of sampling points, N represents the number of data, Y (n) and Y (n−k) represent pixel data, and k represents the order of shifting time. It is assumed that the autocorrelation coefficient value takes an absolute value.

  The comparator 5 calculates the autocorrelation coefficient value of the target pixel from the equation (1), and compares the autocorrelation function values between the frames. If the calculated autocorrelation coefficient value (normalized) is less than 0.5 (autocorrelation coefficient value <0.5), it is determined that there is no correlation between target frames, and if it is 0.5 or more (self-relationship) (Numerical value ≧ 0.5) is determined to be correlated. The correlation determination result between the three frames can be divided into four patterns as shown in FIG. For each of these four patterns, motion determination is performed, and whether or not 3D-DNR is to be applied is determined as follows.

(1) Pattern # 1: This is a case where the correlation between the pixels in the first frame, the second frame, and the third frame is high. In this case, it is regarded as a pixel of a still image, and it is determined that 3D-DNR is applied to the target pixel.
(2) Pattern # 2: A case where the correlation between the first frame and the second frame is high and the correlation between the second frame and the third frame is low. In this case, it is considered that the image starts moving, and it is determined that 3D-DNR is not applied to the target pixel.
(3) Pattern # 3: This is a case where the correlation between the second frame and the third frame is high and the correlation between the first frame and the second frame is low. In this case, it is considered that the motion of the image is stopped, and it is determined that 3D-DNR is applied to the target pixel.
(4) Pattern # 4: This is a case where the correlation coefficient varies between frames 1, 2, and 3, and the correlation is low. In this case, it is regarded as a moving image, and it is determined that 3D-DNR is not applied to the target pixel.

The 3D-DNR unit 6 performs pixel-by-pixel or pixel-based processing based on the image data read from the FRBs 2, 3, and 4 in accordance with the determination results in the correlation determination pattern classification (1) to (4) in the comparator 5. The 3D-DNR process is adaptively performed for each set. Any frame may be used for the 3D-DNR processing, but frame 2 is preferable. When the above processing is completed, the data selector 1 overwrites the FRB storing the image data of the oldest frame with the image data of the next frame. The comparator 5 and the 3D-DNR unit 6 execute a new motion determination and a series of 3D-DNR processes on the three frames of image data in which the overwritten image data is the latest image data.

  As described above, the correlation between the pixels of interest is determined by calculating the autocorrelation coefficient value between the pixels of interest using the high degree of correlation between the pixels constituting the three-frame image. By determining the motion of pixels between frames and selectively multiplying a still image portion in one frame only by 3D-DNR, it is not necessary to multiply an extra 3D-DNR on a fast-moving image. It is possible to perform appropriate 3D-DNR processing by retaining local details of the image, removing noise while maintaining the sharpness of the image after 3D-DNR processing. In addition, since 3D-DNR is performed only on necessary portions of one image, the processing speed can be increased. Furthermore, since the motion determination is performed using the correlation between the images of the three frames, it is possible to perform the 3D-DNR process by capturing an image that moves delicately, such as a motion toward a stop.

  Next, as a second embodiment of the present invention, a specific configuration example of the comparator 5 will be described. FIG. 4 is a block diagram showing the electrical configuration of the comparator 5 of this embodiment. The overall configuration of this embodiment is the same as that of the first embodiment and can be easily analogized. Therefore, the description of the configuration, operation, and effects other than the comparator 5 will be omitted, and only the configuration and operation of the comparator 5 will be described.

  The comparator 5 of this embodiment calculates the autocorrelation coefficient value between frames by performing calculation using the autocorrelation function represented by the expression (1) of Embodiment 1 for each pixel at the same position between three frames. More motion determination is performed, and it is determined from this motion determination result whether or not 3D-DNR processing is performed on the target pixel. Therefore, the comparator 5 reads out the image data from the FRBs 2, 3, and 4 and performs predetermined multiplication, an adder 54 that adds the multiplication results, and an addition result 3 A divider 55 for obtaining a correlation coefficient value between frames by dividing by the above and a motion determination from the calculated correlation coefficient value between frames to determine which one of the four patterns shown in the first embodiment belongs. The correlation determination unit 56 determines whether or not 3D-DNR processing is performed on the target pixel.

  Assume that the FRB 2, 3, 4 store the image data FR 0 of the frame 1, the image data FR 1 of the frame 2, and the image data FR 2 of the frame 3 sorted in time order. FIGS. 5A, 5B, and 5C show examples of target pixel data. The pixel data of k = 0 in FIG. 5A is the original data of the first frame, the pixel data of k = 1 in FIG. 5B is that of the second frame shifted by one frame from the original data, FIG. The pixel data of k = 2 in (c) is the third frame shifted by 2 frames from the original data. n represents the number of sampling points.

  FIG. 6 shows numerical values of the image data in the cases of FIGS. 5 (a), (b), and (c). Yn is pixel data of k = 0, Yn-1 is pixel data of k = 1, Yn-2 is pixel data of k = 2, and Yn is Y (n) in the expression (1) of the first embodiment. Similarly, Yn-1 represents Y (n-1) and Yn-2 represents Y (n-2). Referring to FIG. 5A, Yn is 0.5 when n = 0, 1 when n = 1, and 0.25 when n = 2. Here, referring to the pixel data of each frame at the same time (n = 0), 0.5 is frame 1 (FR0), 1 is frame 2 (FR1), and 0.25. Is frame 3 (FR2). Referring to FIG. 5B, Yn-1 is 0.25 when n = 0, 0.5 when n = 1, and 1 when n = 2. Here, referring to the pixel data of each frame at the same time (n = 0), 0.25 is frame 3 (FR2), and 0.5 is frame 1 (FR0). Is frame 2 (FR1). Further, referring to FIG. 5C, Yn-2 is 1 when n = 0, 0.25 when n = 1, and 0.5 when n = 2. Here, referring to the pixel data of each frame at the same time (n = 0), 1 is frame 2 (FR1), 0.25 is frame 3 (FR2), and 0.5. Is frame 1 (FR0).

  FIG. 7 shows a calculation example of the autocorrelation coefficient value between frames using the above numerical values and the equation (1). First, in order to obtain R (1) when k = 1, that is, an autocorrelation coefficient value between frames 1 and 2, the multiplier 55 calculates YnYn-1. Referring to FIG. 5, when n = 0, a pixel at the same position as the target pixel of the image data FR0 and FR2 may be read and multiplied by FR0 * FR2. Similarly, when n = 1, the pixel at the same position as the target pixel of the image data FR1 and FR0 is read and multiplied by FR1 * FR0. When n = 2, the target pixel of the image data FR2 and FR1 It suffices to read out pixels at the same position and perform multiplication of FR2 * FR1. In the case of k = 1, the adder 54 adds these multiplication results by the adder 54, and makes the addition result 1 / N, that is, 1/3 by the divider 55, so that R (1), The autocorrelation coefficient value between frame 1 and frame 2 can be obtained. Here, R (2) with k = 2, that is, the autocorrelation coefficient value between frame 1 and frame 3 can be similarly obtained from YnYn−2. In the example of FIG. 7, since the autocorrelation coefficient values R (1) and R (2) are less than 0.5, it is determined that there is no correlation.

  Next, an operation example of the correlation determination unit 56 will be described. As described above, the autocorrelation coefficient value between frame 1 and frame 2 is obtained in the first process. In the next processing, the next frame 4 is stored in the FRB storing the frame 1, and the autocorrelation coefficient value between the frames 2 and 3 is obtained from the image data of the frames 3 and 4 with the frame 2 as the original data. It is done. Thereby, the motion determination between the frames 1 and 2 and between the frames 2 and 3 is performed. Based on the result of the motion determination, it is determined which of the four patterns # 1 to # 4 shown in FIG. 1 (1) to (4), it is determined whether or not 3D-DNR is applied, and the determination result is notified to the 3D-DNR unit.

  In the above-described operation example of the correlation determination unit, the example in which the correlation determination is performed in two processes has been described. However, in the first process, the autocorrelation coefficient values R (1) and R (2) are obtained, To determine the motion between the frames 1 and 2 and between the frames 1 and 3, and from the result of the motion determination, divide the pattern into four patterns different from those in FIG. 3 is as high as 0.5 or more, or the autocorrelation coefficient between frames 1 and 2 is low, less than 0.5, but the autocorrelation coefficient between frames 1 and 3 is 0. If it is higher than 5, it may be determined that 3D-DNR is applied to the pixel of the frame 1.

  Next explained is the third embodiment of the invention. In the first and second embodiments, an autocorrelation coefficient value is obtained for a pixel having the same coordinates as the target pixel of the image data for three frames, and a motion is determined from the autocorrelation coefficient value. 3D-DNR is adaptively performed on the pixels that are determined to be. On the other hand, in this embodiment, the autocorrelation coefficient value is obtained for a pixel having the same coordinates as the target pixel of the image data for three frames, and the autocorrelation coefficient value is also calculated between the peripheral pixels of the target pixel in the preceding and following frames. Calculate the average value of the autocorrelation coefficient value between the pixels with the same coordinates as the target pixel and the autocorrelation coefficient value between the surrounding pixels of the target pixel as the autocorrelation coefficient value between frames, and determine the motion from this autocorrelation coefficient value. And 3D-DNR is adaptively performed on the pixel determined to be still or stopped from the motion determination result.

  The basic configuration of this embodiment is the same as that shown in FIG. The different part is the configuration and operation of the comparator 5. Therefore, since the configuration and operation similar to those of the first embodiment can be easily analogized, their description is omitted, and only the configuration and operation of the comparator 5 will be described.

  Similar to the first embodiment, the comparator 5 of this embodiment reads out pixels having the same coordinates as the target pixel of the image data for three frames from the FRBs 2, 3, and 4 and obtains an autocorrelation coefficient value. Subsequently, the target pixel in the middle frame and the peripheral pixels of the target pixel in the previous and subsequent frames are read from the FRBs 2, 3, and 4, and the autocorrelation coefficient value is obtained. The relational value and the average value of the autocorrelation coefficient values between the target pixel and the peripheral pixels obtained subsequently are calculated as autocorrelation coefficient values between frames. Next, a correlation determination is performed from the autocorrelation coefficient value, and from this correlation determination result, it is determined which of the patterns in FIG. 3 belongs, a motion determination is performed, and the motion determination result is a stationary or motion stop pattern. It is determined that 3D-DNR is to be performed on the pixel determined to be.

  Next, the case where this embodiment is applied to Embodiment 2 will be described. The comparator 5 of this embodiment is configured in the same manner as in the embodiment 2, and first calculates the autocorrelation coefficient value between the target pixels of the frames 1 and 2. Next, the peripheral pixel data of the target pixel of frame 1 and the target pixel data of frames 2 and 3 are read from the FRB, and the autocorrelation coefficient value between the peripheral pixel of the target pixel of frame 1 and the target pixel of frame 2 is calculated. The average value of the autocorrelation coefficient value between the pixel of interest of frame 1 and frame 2 previously obtained and the autocorrelation coefficient value between the peripheral pixel of frame 1 obtained next and the pixel of interest of frame 2 is obtained as frames 1 and 2. The autocorrelation coefficient value between. Subsequently, as in the second embodiment, the autocorrelation coefficient value between the target pixels of the frames 2 and 3 is calculated based on the updated image data of the frames 2, 3, and 4. Next, the pixel-of-interest pixel data of frame 2 and the pixel-of-interest pixel data of frames 3 and 4 are read from the FRB, and the autocorrelation coefficient value between the pixel of interest of frame 2 and the pixel of interest of frame 3 is calculated. The average value of the autocorrelation coefficient value between the pixel of interest of the frame 2 and the frame 3 previously obtained and the autocorrelation coefficient value between the pixel of interest of the frame 2 and the peripheral pixel of the pixel of interest of the frame 3 obtained next, The autocorrelation coefficient value between frames 2 and 3 is used.

  The correlation determination unit 56 of the comparator 5 determines between the frames 1 and 2 and between the frames 2 and 3 based on the autocorrelation coefficient value between the frames 1 and 2 and the autocorrelation coefficient value between the frames 2 and 3 calculated above. A motion determination is performed, and from the result of the motion determination, it is determined which of the four patterns # 1 to # 4 shown in FIG. 3 belongs, and as shown in (1) to (4) of the first embodiment, It is determined whether or not 3D-DNR is applied, and the determination result is notified to the 3D-DNR unit.

  As in this embodiment, when 3D-DNR is applied in consideration of not only the correlation between the target pixels but also the correlation with the surrounding pixels, it is possible to discriminate between noise and an image. Therefore, it becomes possible to easily remove the fixed beat noise that could not be removed conventionally.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and the present invention can be changed even if there is a design change or the like without departing from the gist of the present invention. include. For example, in Example 3 described above, in order to reflect the autocorrelation coefficient value between the surrounding pixels in the autocorrelation coefficient value between the target pixels, the average value of both is used as the autocorrelation coefficient value between the target frames. Alternatively, it may be reflected by another calculation method such as a weighted average value. In each embodiment, the comparator 5 has been described as a hardware configuration. However, the comparator 5 is configured by a computer, and each functional component of the comparator 5 described with reference to FIG. 1 or 4 is realized by a program of the computer. It is also possible.

  The present invention is not limited to a device that needs to write a large amount of image data in a storage means such as a memory cell, and is suitable for a television receiver, a video recorder, and the like because 3D-DNR processing is fast, for example. Is available.

1 is a block diagram showing an electrical configuration of an image processing apparatus according to a first embodiment of the present invention. It is a figure explaining operation | movement of the image processing apparatus. It is the figure which divided the autocorrelation between three frames in the Example into patterns. It is a block diagram which shows the electric constitution of the comparator in the image processing apparatus which is 2nd Example of this invention. (A), (b), (c) is a figure which shows an example of the attention pixel data in the Example. It is the figure which showed each image data of Fig.5 (a), (b), (c) numerically. It is a figure which shows the example of calculation of the autocorrelation coefficient value between frames in the Example.

Explanation of symbols

1 Data selector (image selection means)
2 Frame buffer (memory means)
3 Frame buffer (memory means)
4 Frame buffer (memory means)
5 Comparator (Judgment means)
6 3D-DNR unit (image processing means)

Claims (4)

  An image processing apparatus for performing three-dimensional noise reduction processing on image data, wherein image selection means for distributing input image data to a storage means for storing image data for three frames for each frame, and image data between the frames Calculation means for calculating an autocorrelation coefficient value for each pixel, determination means for determining whether to perform a three-dimensional noise reduction process for each pixel based on the calculated autocorrelation coefficient value, and a three-dimensional noise reduction process. An image processing apparatus comprising image processing means for adaptively performing a three-dimensional noise reduction process on a pixel or a set of pixels determined to be performed.   The determination unit determines an image motion from the calculated autocorrelation coefficient value, and determines a pixel determined to be still or stopped from the image motion determination result as a pixel to be subjected to three-dimensional noise reduction processing. The image processing apparatus according to claim 1, wherein the image processing apparatus is an image processing apparatus.   The determination means performs a motion determination by reflecting an autocorrelation coefficient value between a target pixel of an arbitrary frame and peripheral pixels of another frame in an autocorrelation coefficient value between the target pixels, and determines a motion from the result of the motion determination. The image processing apparatus according to claim 1, wherein a pixel determined to stop moving is determined to be a pixel that performs a three-dimensional noise reduction process.   When the determination means determines that the autocorrelation coefficient values between an arbitrary frame and the two frames before and after the arbitrary frame are correlated, it determines that the target pixel is stationary, and the arbitrary frame If the autocorrelation coefficient value between the frame and the previous frame is determined to be uncorrelated and the autocorrelation coefficient value between the arbitrary frame and the subsequent frame is determined to be correlated, 4. The image according to claim 2, wherein it is determined that the pixel of interest has stopped moving, and the pixel that has been determined to be stationary or stopped is determined to be a pixel that performs three-dimensional noise reduction processing. Processing equipment.

Images