JP4930541B2 - Image processing method and image processing apparatus - Google Patents

Description

  The present invention relates to an image processing method and an image processing apparatus that can be applied to an imaging apparatus and perform noise reduction.

  In recent years, shooting with a lower illuminance or shooting with a shorter exposure time has been demanded when shooting with a camera or video. However, under low illuminance and a short exposure time, the amount of incident light is small and the amount of accumulated charge is reduced, so that it is necessary to amplify the signal in order to obtain a bright image. At that time, noise generated in the image pickup device or the analog circuit is also amplified, and the obtained image has a low image quality with a lot of noise. As an improvement measure for this problem, it is possible to improve the S / N by increasing the charge accumulation amount by performing long-time exposure and relatively lowering the random component. However, long-time exposure causes image blur due to camera shake during exposure (referred to as exposure blur as appropriate), so the imaging device must be fixed by a tripod or the like.

  As another method, noise that is uncorrelated in the time direction is obtained by repeatedly capturing images at a shutter speed that hardly causes exposure blur, for example, 1/30 second, and averaging a plurality of images obtained by continuous shooting. There are ways to reduce this. However, there is a problem that degradation such as blurring of an image or multiple exposure-like dragging of a moving subject occurs due to camera movement due to camera shake or the like and a moving subject.

  Furthermore, Japanese Patent Application Laid-Open No. H10-228561 describes that a plurality of images obtained by continuous shooting are corrected and superimposed. Here, the camera shake is a camera shake that occurs between exposure periods, for example, one field and one frame. Similarly, Patent Document 2 describes that an image with good image quality is obtained by dividing the total exposure time into a plurality of exposure periods and correcting and adding the images obtained in each exposure period. .

JP-A-9-261526

Japanese Patent Laid-Open No. 11-75105

  As described in Patent Document 1 or Patent Document 2, the method of adding a plurality of motion-corrected images is based on the accuracy of a detected motion vector in an image in which an unnecessary moving subject or the like has entered and a low-contrast image. However, there is a drawback that the image quality is not sufficiently improved. Further, the accuracy of the process of adding with a weighting coefficient in units of one frame is insufficient. Furthermore, since the noise increases as the exposure time becomes longer, the amount of noise may differ depending on the shutter speed even with the same exposure amount. Unless such shooting conditions are taken into consideration, noise cannot be removed with high accuracy.

  Accordingly, an object of the present invention is to provide an image processing method and an image processing apparatus capable of removing noise with high accuracy.

In order to solve the above-described problem, the present invention is an image processing method in which an image with reduced noise is output using first and second still images that are temporally continuous,
Performing block-based motion compensation on the second still image;
Generating a table for setting an addition weight for a pixel difference absolute value between the first still image and the second still image subjected to motion compensation, and setting the addition weight using the table ;
The first still image and a second still image subjected to motion compensation, possess and obtaining an image after addition or weighted average processing according to the weighting factor,
This is an image processing method in which the relationship of the amount of noise corresponding to the pixel value under each shooting condition of a plurality of shooting conditions is held in advance, and an addition weight is set by defining a table according to the current shooting conditions .
The present invention is an image processing method for generating a noise-reduced image using first, second and third still images that are temporally continuous,
Performing block-based motion compensation on the second still image;
Generating a table for setting the first addition weight for the pixel difference absolute value between the first still image and the second still image subjected to motion compensation, and setting the first addition weight using the table ;
The first still image and the second still image that has undergone motion compensation are added or weighted averaged according to the first addition weight to obtain a processed image, and block-based motion compensation is performed on the third still image. Steps to perform,
Generating a table for setting the second addition weight for the pixel difference absolute value between the first still image and the third still image subjected to motion compensation, and setting the second addition weight using the table ;
The third still picture that has been subjected to the predetermined image and the motion compensation after treatment, addition or weighted average to possess and obtaining an image after processing in accordance with a second weighting factor,
The relationship between the noise amounts corresponding to the pixel values under each of the plurality of shooting conditions is held in advance, and a first addition weight and a second addition weight are set by defining a table according to the current shooting conditions. This is an image processing method.

The present invention is an image processing method for performing noise reduction on a moving image input at unit time intervals,
Performing block-unit motion compensation on an image after processing one unit time ago;
Generating a table for setting an addition weight for a pixel difference absolute value between the current moving image and an image after processing one unit time before performing motion compensation, and setting the addition weight using the table ;
An image after one unit time before the process was the current video and motion compensation, by adding or weighted average according to the weighting factor, have a obtaining a processed image corresponding to the current video,
This is an image processing method in which the relationship of the amount of noise corresponding to the pixel value under each shooting condition of a plurality of shooting conditions is held in advance, and an addition weight is set by defining a table according to the current shooting conditions .
The present invention is an image processing method for performing noise reduction on a moving image input at unit time intervals,
A motion vector in block units is obtained from the current moving image and the image after processing one unit time ago, and the motion vector is obtained for the image after processing one unit time ago and the image after motion compensation two units time ago. Performing motion compensation;
A table for setting the first addition weight is generated for the pixel difference absolute value between the current moving image and the image after processing one unit time before motion compensation is performed, and the first addition weight is set using the table. And steps to
Adding the current video and the processed image one unit time before performing motion compensation in accordance with a first addition weight to obtain an image after the first processing;
A table for setting the second addition weight is generated for the pixel difference absolute value between the motion-compensated image after motion compensation 2 units before and the image after the first processing. Setting an addition weight;
The motion-compensated image after motion compensation 2 units before and the image after the first processing are added or weighted averaged according to the second addition weight, and after the second processing corresponding to the current moving image The step of obtaining an image and
The relationship between the noise amounts corresponding to the pixel values under each of the plurality of shooting conditions is held in advance, and a first addition weight and a second addition weight are set by defining a table according to the current shooting conditions. This is an image processing method.

The present invention is an image processing apparatus that generates a noise-reduced image using first and second still images that are temporally continuous,
A motion compensation device that performs motion compensation on a block basis for the second still image;
A determination device that generates an addition weight for a pixel difference absolute value between the first still image and the second still image subjected to motion compensation, and sets the addition weight using the table ;
A second still image was first still and motion compensation, possess an adding device for obtaining an image after addition or weighted average processing according to the weighting factor,
This is an image processing apparatus that preliminarily stores a relationship between noise amounts corresponding to pixel values in each of a plurality of shooting conditions, and sets an addition weight by defining a table according to the current shooting conditions .
The present invention is an image processing apparatus that generates an image with reduced noise using first, second, and third still images that are temporally continuous,
A motion compensation device, a determination device, and an addition device;
The motion compensation device performs motion compensation for each block on the second still image,
The determination device generates a table for setting the first addition weight for the pixel difference absolute value between the first still image and the second still image subjected to motion compensation, and sets the first addition weight using the table. And
The addition device performs addition or weighted averaging on the first still image and the second still image subjected to motion compensation according to the first addition weight to obtain a processed image,
The motion compensation device performs motion compensation for each block on the third still image,
The determination device generates a table for setting the second addition weight for the pixel difference absolute value between the first still image, the processed image, and the third still image subjected to motion compensation. Set 2 addition weights,
By the addition device, the processed image and the third still image subjected to motion compensation are added or weighted averaged according to the second addition weight to obtain a processed image ,
The relationship between the noise amounts corresponding to the pixel values under each of the plurality of shooting conditions is held in advance, and a first addition weight and a second addition weight are set by defining a table according to the current shooting conditions. An image processing apparatus configured as described above.

The present invention is an image processing apparatus that performs noise reduction on a moving image input at unit time intervals,
A motion compensation device, a determination device, and an addition device;
A motion compensation device generates a table for setting an addition weight for the pixel difference absolute value of the image after processing one unit time ago, and performs block-unit motion compensation on the table ,
The determination device generates a table for setting the addition weight for the absolute value of the pixel difference between the current moving image and the image after processing one unit time before the motion compensation, and sets the addition weight using the table. And
By the addition device, the processed image corresponding to the current moving image is obtained by adding or weighted average the current moving image and the processed image one unit time before performing motion compensation according to the addition weight ,
An image processing apparatus in which a relationship between noise amounts corresponding to pixel values in each shooting condition of a plurality of shooting conditions is held in advance, and an addition weight is set by defining a table according to the current shooting conditions is there.
The present invention is an image processing apparatus that performs noise reduction on a moving image input at unit time intervals,
A motion compensation device, a determination device, and an addition device;
A motion vector in block units is obtained from the current moving image and the image after processing one unit time ago, and the motion vector is obtained for the image after processing one unit time ago and the image after motion compensation two units time ago. Motion compensation,
Generate a table for setting the first addition weight for the pixel difference absolute value between the current moving image and the processed image one unit time before the motion compensation , and use the table to generate a noise amount based on the shooting conditions And set the first addition weight using
The current image and the processed image one unit time before performing the motion compensation are added or weighted average according to the first addition weight to obtain the first processed image,
A table for setting the second addition weight is generated for the pixel difference absolute value between the motion-compensated image after motion compensation 2 units before and the image after the first processing. Set the addition weight,
The motion-compensated image after motion compensation 2 units before and the image after the first processing are added or weighted averaged according to the second addition weight, and after the second processing corresponding to the current moving image Get an image ,
The relationship between the noise amounts corresponding to the pixel values under each of the plurality of shooting conditions is held in advance, and a first addition weight and a second addition weight are set by defining a table according to the current shooting conditions. An image processing apparatus configured as described above.

  According to the present invention, it is possible to obtain a noise reduction effect even with respect to movement of a camera or a moving subject during shooting under low illuminance. In the present invention, highly accurate noise reduction can be performed by performing addition determination in units of pixels. In addition, occurrence of block distortion can be suppressed, and color noise of a large size generated in the flat portion can be reduced.

  Hereinafter, an example of an imaging apparatus including an image processing apparatus according to an embodiment of the present invention will be described. The imaging device will be described with reference to the drawings. The overall configuration of one embodiment is shown in FIG. Subject light is incident on the image sensor 101 via an imaging optical system (not shown).

An imaging optical system includes a zoom lens for enlarging / reducing an object, a focus lens for adjusting a focal length, an iris (aperture) for adjusting a light amount, an ND (Neutral Density) filter, and these lenses and A drive circuit for driving the iris is provided. Although not shown, the zoom lens, focus lens, IRIS, and ND filter are driven by a driver.

The image sensor 101 is a CCD (Charge Coupled Device), a CMOS (Complementary Metal Oxide Semiconductor), or the like. An imaging signal corresponding to subject light is captured by the imaging element 101. An example of the imaging device is a camcorder (meaning a camera and a recorder) capable of shooting a moving image and a still image, but may be a still image digital camera, a PDA (Personal Digital Assistants), a mobile phone, or the like.

The image sensor 101 may be either a primary color system or a complementary color system, and is sampled and held for each RGB primary color signal obtained by photoelectrically converting an incident subject image or a color signal of a complementary color image signal, and AGC (Automatic Gain Control). ) To control the gain and output an imaging signal (referred to as row data) converted to a digital signal by A / D conversion. Raw data from the image sensor 101 is supplied to a camera signal processing unit 102 configured as an IC (Integrated Circuit). In the camera signal processing unit 102, development processing and conversion processing into luminance and color difference signals are performed.

  An input image signal from the camera signal processing unit 102 is supplied to the noise reduction device 103, and noise reduction processing is performed. From the noise reduction device 103, an output image signal with reduced noise is obtained. Imaging condition information used for noise amount estimation is supplied to the noise reduction apparatus 103 from a camera control unit (configured by a microcomputer) 104. The shooting condition information includes a shooting mode, an exposure time, a gain, and the like.

  FIG. 2 shows an example of the noise reduction apparatus 103. The noise reduction apparatus 103 includes a frame memory 201, a motion prediction apparatus 202, a motion compensation apparatus 203, an addition determination apparatus 204, an addition apparatus 205, and a hidden area processing apparatus 206.

  Each component of the noise reduction apparatus 103 will be described. A standard image and a reference image are input to the motion prediction device 202. The motion prediction apparatus 202 aligns the reference image input in block units designated by the image size with respect to the standard image, and outputs a motion vector MV in block units. The motion vector MV indicates the amount and direction of movement of the reference image with respect to the base image. For example, the motion vector MV can be detected by a block matching method. As the block size, for example, a macro block can be used.

  The motion compensation device 203 performs motion compensation of an input image according to the motion vector MV, and outputs a motion compensated image (referred to as an MC image as appropriate). At this time, the input image may be switched by the switch SW with respect to the motion vector MV obtained between a pair of images, and motion compensation of the image selected by the switch SW may be performed. The switch SW selects one of the reference image and the MC image of the previous frame and inputs the selected image to the motion compensation device 203.

  The MC image is supplied to the addition determination device 204 and the addition device 205. The reference image is supplied to the addition determination device 204 and the addition device 205. Furthermore, shooting condition information from the camera control unit is supplied to the addition determination device 204. The addition device 205 performs weighted addition using the addition weight formed by the addition determination device 204.

  The present invention can be applied to noise reduction of both still images and moving images. For example, during still image shooting, a plurality of still images are continuously shot. Continuous shooting is performed at the fastest continuous shooting speed of the imaging apparatus. When the image sensor 101 is a CCD, for example, continuous shooting is performed at a speed of about 2 sheets per second, and when the image sensor 101 is a CMOS, for example, continuous shooting is performed at a speed of about 30 to 60 sheets per second. In this case, the first photographed image is the reference image, the second photographed image is the reference image, the reference image is motion compensated with the motion vector MV, and the MC image is formed. The MC image and the MC image are added by the addition device 205 using the addition weight from the addition determination device 204 to form an image with reduced noise (referred to as an NR image as appropriate).

  The first NR image output from the adding device 205 is stored in the frame memory 201. The first NR image is treated as a standard image, the second captured image is treated as a reference image, and the noise reduction operation similar to that described above is performed by these two images. By repeating the noise reduction operation, the random noise component is reduced.

  The reference image and the MC image are supplied to the hidden area processing device 206. The hidden area processing device 206 suppresses the propagation of image quality degradation of the MC part of the hidden area with low reliability when using the MC image that is one or more images away from the standard image as a reference image. Is provided. The processed image is stored in the frame memory 201.

  The addition weight used by the adder 205 is formed by the adder determiner 204 on a pixel-by-pixel basis. The addition weight represents the reliability of addition. That is, the correctness of motion compensation is determined between the reference image and the MC image, and the pixel that reduces noise without causing deterioration when the reference image and the MC image are added has high reliability. And Specifically, when pixels obtained by photographing the same part of the same object in images taken at different times are added, noise uncorrelated to time can be reduced by the addition, and thus it can be said that the reliability is high. In other words, the reliability indicates whether noise is superimposed on the same image or whether noise is superimposed on different images, and is added corresponding to only the amount of noise as much as possible. The formation of weights is required for highly accurate noise reduction processing. The addition weight has a value of 0 when the reliability is low and 1 when the reliability is high. As will be described later, the weighted average of the reference image and the MC image is calculated by the adding device 205. The weighted average weight is a value obtained by multiplying the addition weight by the feedback rate. For example, if the feedback rate is set to 1/3, when the addition weight changes from 0 to 1, the weight for the MC image in the weighted average changes from 0 to 1/3.

  An example of the addition determination device 204 is shown in FIG. In the addition determination device 204, a similar determination unit is provided for each of the luminance signal Y and the two color difference signals Cb and Cr. First, the configuration of the determination unit related to the luminance signal Y will be described.

  Pixel data at the same spatial position of the reference image and the MC image is supplied to the subtracter 302 via the pre-filters 300 and 301, the difference between the pixel values is calculated, and converted into an absolute value by the absolute value calculator 303. The The pre-filters 300 and 301 cut high frequency components in the image and improve the effect of addition determination. The absolute value calculator 303 outputs the absolute value of the input signal. The absolute value of the difference is supplied to the pixel addition determiner 304.

  An output signal of the noise amount estimator 305 is supplied to the pixel addition determiner 304. The noise amount estimator 305 estimates the amount of noise corresponding to the pixel value, and inputs shooting condition information such as a shooting mode, exposure time, and gain information that affects the noise amount in the image from the outside. The data of the relationship between the luminance value measured in step 1 and the noise amount is included as a table. From this table, the relationship between the luminance value of the corresponding condition and the noise amount is obtained, and the noise amount corresponding to the input luminance value is output.

  For example, when the exposure time becomes longer, random noise tends to increase. Further, when the gain for amplifying the imaging signal is increased, the random noise tends to increase. Therefore, the shutter speed and / or gain information is used as shooting conditions, and a table of pixel values and noise amounts corresponding to each of the shutter speed and / or gain is prepared and prepared in advance. Furthermore, a shooting mode provided in the imaging apparatus may be used as a shooting condition. Note that shooting conditions are set for each frame, for example.

  FIG. 4 conceptually shows a pixel value versus noise amount table provided in the noise amount estimator 305. The relationship between the pixel value and the amount of noise is defined for each of the different shooting conditions 1 to N. The noise amount estimator 305 inputs a shooting mode that affects the amount of noise in an image, a shutter speed (exposure time), a gain, and the like as shooting condition information from the camera control unit 104, and is a luminance value measured in advance under each shooting condition. The relationship between the luminance value of the corresponding condition and the noise amount is obtained from the relationship between the noise amount and the noise amount, and the noise amount corresponding to the input luminance value under the corresponding photographing condition is output.

  The noise amount from the noise amount estimator 305 and the absolute difference value of the pixel value (luminance) from the absolute value calculator 303 are supplied to the pixel addition determiner 304. The absolute difference value is an input index value for the pixel addition determination unit 304. The pixel addition determination unit 304 outputs an addition weight that is the reliability of addition corresponding to the index value.

  FIG. 5 shows an example of the pixel addition determiner. The luminance pixel addition determination unit 304 includes a converter 601, a determination curve generator 602, and an addition determination unit 603 that convert the noise amount from the noise amount estimator 305 into a determination threshold value. The converter 601 converts the noise amount into a determination threshold using a table determined based on the visual characteristics, and outputs the determination threshold. The shape of the determination curve is determined according to the determination threshold. The determination curve represents the shape of a determination curve representing the correspondence between the input index value representing the error amount and the addition weight representing the reliability with respect to the error.

  FIG. 6 shows some examples of determination curves that can be used. The horizontal axis of the graph showing each curve is the input index value, and the vertical axis indicates the addition weight for the index value. In each curve, an index value value at which the addition weight changes from 1 and an index value value at which the addition weight is 0 are determination threshold values. The magnitude of the determination threshold is a value corresponding to the amount of noise. When the amount of noise is large, the determination threshold is set to a larger index value.

  The determination curve generator 602 determines and outputs a determination curve using the determination threshold value that is the output of the converter 601. The determination curve is configured as a table, for example, and stored in the memory. The addition determination unit 603 inputs an index value using this determination curve, and outputs an addition weight corresponding to the index value. This addition weight is supplied from the pixel addition determination unit 304 to the multiplier 306. The series of conversion, calculation, and determination of the addition determination unit 304 is performed in units of pixels because a determination curve generated using a noise amount corresponding to the pixel value is used in units of pixels.

  In the example of the determination curve shown in FIG. 7, the output (addition weight to the index value) is 1 for the index value up to the determination threshold 1, and the index value is between the determination threshold 1 and the determination threshold 2. Thus, the addition weight gradually decreases toward zero. The determination threshold 1 and the determination threshold 2 define the shape of the determination curve. When the amount of noise is large, the index values corresponding to the determination threshold value 1 and the determination threshold value 2 are large. By the determination curve shown in FIG. 7, the input index value is converted into a corresponding addition weight and output.

  The addition weight output from the addition determination unit 304 is supplied to the multiplier 306, and the moving average of the luminance signal is multiplied by the addition weight obtained in the same manner as the above-described pixel difference absolute value. The output of the multiplier 306 is supplied to the multiplier 325 and multiplied by the addition weight obtained in the same manner as the pixel difference absolute value described above with respect to the moving variance of the luminance signal. Further, the output of the multiplier 325 is supplied to the multiplier 326, and is multiplied by the addition weight obtained with respect to the pixel, moving average, and moving variance of the color signal. The addition weight generated by the addition determination device 204 is extracted from the output of the multiplier 326. This addition weight is supplied to the addition device 205.

  Next, the case where the index value is the moving average difference absolute value and the moving variance difference absolute value of the luminance signal will be described. As shown in FIG. 3, pixel data at the same spatial position of the reference image and the MC image are supplied to the moving average calculators 310 and 311 via the pre-filters 300 and 301. Similarly, pixel data at the same spatial position of the reference image and the MC image are supplied to the moving dispersion calculators 320 and 321 via the prefilters 300 and 301.

  The moving average calculators 310 and 311 calculate a moving average of pixel values in a rectangular area centered on a pixel to be calculated (center pixel), for example, an area of (5 × 5) pixels. In accordance with the movement of the center pixel, the moving area is calculated by shifting the rectangular area while overlapping the calculation area of the previous pixel. The moving variance calculators 320 and 321 calculate the variance of pixel values within a rectangular area centered on the pixel to be calculated (center pixel). According to the movement of the central pixel, the movement distribution is calculated by shifting the rectangular area while overlapping the calculation area of the previous pixel.

  The reason why the addition weight is obtained with respect to the moving average and moving variance is to more accurately reduce the amount of noise. For example, when moving averages or moving variances of multiple pixel areas are different, it is often more correct to determine that the images are different than to determine the absolute difference of pixel values as noise. is there.

  The output signals of the moving average calculators 310 and 311 are supplied to the subtractor 312, the output signals of the moving dispersion calculators 320 and 321 are supplied to the subtractor 322, and the difference of the moving average and the difference of the moving variance are respectively calculated. The absolute value calculators 313 and 323 convert the absolute value. The absolute value calculator 313 outputs the absolute value of the moving average difference signal, and the absolute value calculator 323 outputs the absolute value of the moving variance difference signal. The absolute value of the difference is supplied to the moving average addition determiner 314 and the moving variance addition determiner 324, respectively.

  The output signal of the noise amount estimator 305 is supplied to the moving average addition determiner 314 and the moving variance addition determiner 324. The moving average addition determiner 314 and the moving variance addition determiner 324 have the same configuration as the pixel addition determiner 304 (see FIG. 5), respectively, and generate a determination curve (table) based on a determination threshold obtained by converting the amount of noise. An addition weight is obtained by giving a moving average difference absolute value and a moving variance difference absolute value as index values of the determination curve. The addition weight from the moving average addition determination unit 314 for the luminance signal is supplied to the multiplier 306, and the addition weight from the moving dispersion addition determination unit 324 for the luminance signal is supplied to the multiplier 325.

  The above is the processing for obtaining the addition weight using the pixel difference absolute value, moving average difference absolute value, and moving variance difference absolute value of the luminance signal as index values. Similarly, addition weights are obtained for color signals. In the case of a component color video signal, as in (4: 2: 0), (4: 2: 2), (4: 1: 1), a sampling frequency related to the luminance signal Y, a sampling frequency related to the color difference signal Cb, The ratio with the sampling frequency regarding the color difference signal Cr is represented. For each of the color difference signals Cb and Cr, an addition weight is obtained in the same manner as the luminance signal Y described above.

  The luminance signal and the color difference signals Cb and Cr are subjected to addition weight generation processing for the same region on the image. As can be seen from the ratio of the sampling frequencies described above, the number of addition weights obtained for the color difference signals Cr and Cb is insufficient compared to the addition weight obtained for the luminance signal Y. In order to solve this problem, an up-sampling circuit 527 that performs interpolation is provided, and the addition weight related to the color difference signal from the up-sampling circuit 527 is supplied to the multiplier 326 and multiplied by the addition weight related to the luminance signal.

  Processing related to the color difference signal Cb will be described. The color difference signal Cb of the reference image is supplied to the noise amount estimator 405 via the prefilter 400. Imaging condition information is supplied to the noise amount estimator 405. The noise amount estimator 405 outputs an estimated value of the noise amount from the noise estimator 405 by using a table indicating the relationship between the pixel value and the noise amount corresponding to each shooting condition, similarly to the noise amount estimator 305 regarding the luminance signal. . The estimated value of the noise amount is output to the pixel addition determination unit 404.

  The color difference signal Cb of the reference image and the color difference signal Cb of the MC image are supplied to the subtractor 402 via the pre-filters 400 and 401, respectively, and the output of the subtractor 402 is converted into an absolute value by the absolute value calculator 403, and the pixel addition The determination unit 404 obtains an addition weight using the pixel difference absolute value as an index value. The pixel addition determination unit 404 has the same configuration as the pixel addition determination unit 304 (see FIG. 5) described above, and the pixel addition determination unit 404 has an addition weight that is the reliability of addition corresponding to the index value (pixel difference absolute value). Is output.

  Next, the case where the index value is the moving average difference absolute value and the moving variance difference absolute value of the color difference signal Cb will be described. Pixel data at the same spatial position of the reference image and the MC image is supplied to the moving average calculators 410 and 411 via the prefilters 400 and 401. Similarly, pixel data at the same spatial position of the reference image and the MC image are supplied to the moving dispersion calculators 420 and 421 via the prefilters 400 and 401. The moving average calculators 410 and 411 calculate a moving average of pixel values in a rectangular area centered on a pixel to be calculated (center pixel). In accordance with the movement of the center pixel, the moving area is calculated by shifting the rectangular area while overlapping the calculation area of the previous pixel. The moving variance calculators 420 and 421 calculate the variance of pixel values within a rectangular area centered on the pixel to be calculated (center pixel). According to the movement of the central pixel, the movement distribution is calculated by shifting the rectangular area while overlapping the calculation area of the previous pixel. The reason why the addition weight is obtained with respect to the moving average and moving variance is to more accurately reduce the amount of noise.

  The output signals of the moving average calculators 410 and 411 are supplied to the subtractor 412, the difference of the moving average is calculated, the output signals of the moving dispersion calculators 420 and 421 are supplied to the subtractor 422, and the difference of the moving dispersion is calculated. The absolute value calculators 413 and 423 convert the absolute values. The output signal of the noise amount estimator 405 is supplied to the moving average addition determiner 414 and the moving variance addition determiner 424. The moving average addition determination unit 414 and the moving dispersion addition determination unit 424 have the same configuration as the pixel addition determination unit 304 (see FIG. 5), and the pixel addition determination units 414 and 424 have the moving average difference absolute value and the moving dispersion difference. Using the absolute values as index values, an addition weight that is the reliability of the addition is output.

  The addition weight from the moving average addition determination unit 414 of the color difference signal Cb is supplied to the multiplier 406 and is multiplied by the addition weight from the addition determination unit 404. The addition weight of the chrominance signal Cb from the moving variance addition determination unit 424 is supplied to the multiplier 425 and multiplied by the output of the multiplier 406.

  The other color difference signals Cr have the same configuration as the color difference signal Cb described above. That is, the pre-filters 500 and 501, the noise amount estimator 505, the subtractor 502, the absolute value calculator 503, and the pixel addition determiner 504 obtain addition weights using the pixel difference absolute value of the color difference signal Cr as an index value. The addition weights using the moving average difference absolute value of the color difference signal Cr as an index value are pre-filters 500 and 501, noise amount estimator 505, moving average calculators 510 and 511, subtractor 512, absolute value calculator 513, moving average. It is obtained by the addition decision unit 514. The addition weights using the moving variance difference absolute value of the chrominance signal Cr as an index value are prefilters 500 and 501, noise amount estimator 505, moving variance calculators 520 and 521, subtractor 522, absolute value calculator 523, moving average. It is obtained by the addition decision unit 524.

  The addition weight using the pixel difference absolute value as the index value and the addition weight using the moving average difference absolute value as the index value are multiplied by the multiplier 506, and the output of the multiplier 506 and the moving variance difference absolute value are used as the index value. Is multiplied by a multiplier 525. The multiplier 526 multiplies the addition weight related to the color difference signal Cr output from the multiplier 525 and the addition weight related to the color difference signal Cb described above. The output of the multiplier 526 is supplied to the upsampling circuit 527. The upsampling circuit 527 interpolates the color difference addition weight according to the color difference format, and aligns the score with the luminance addition weight. The output of the upsampling circuit 527 is multiplied by a multiplier 326. The output of the multiplier 326 is an addition weight indicating the final reliability of the pixel of interest. The addition weight represents the reliability of each of the luminance pixel, the luminance moving average, the luminance moving variance, the chrominance pixel, the chrominance moving average, and the chrominance moving variance, and is represented by continuous values of 0 to 1, respectively.

  An example of the adding device 205 is shown in FIG. A multiplier 605 multiplies the addition weight, which is the output of the addition determination device 204, by w, which is a feedback coefficient for the MC image. A coefficient (represented by α) output from the multiplier 605 is supplied to the multiplier 606 and multiplied by the MC image. The coefficient α is supplied to the subtractor 607, (1-α) is generated, and (1-α) is supplied to the multiplier 608 to be multiplied by the reference image. The outputs of the multipliers 606 and 608 are added by an adder 609, and an NR image that is an image after noise removal is taken out from the adder 609. The feedback coefficient w is, for example, 1 / (k + 1) when the first frame is 0th and the subsequent frames are 1st, 2nd, ..., kth. In this example, when it is determined that noise exists in the same image and the addition weight is the maximum value (1), α = 1/2, and it is determined that the images are different and the addition weight is In the case of the minimum value (0), α = 0.

  An example of the hidden area processing device 206 is shown in FIG. The hidden area processing device 206 suppresses the propagation of image quality degradation of the MC part of the hidden area with low reliability when using the MC image that is one or more images away from the standard image as a reference image. Is provided. As shown in FIG. 9, the subtracter 611 obtains a difference value between the reference image and the MC image, and the subtractor 612 subtracts the threshold value Thr from the difference value. The subtraction result is supplied to the switch controller 613. The pixels included in the hidden area with low reliability are detected as pixels larger than the threshold Thr, that is, the subtraction result is positive.

  The MC image pixel is supplied to one input terminal of the switch 614, and the reference image pixel is supplied to the other input terminal. The switch 614 replaces the pixel having the difference value larger than the threshold value Thr with the pixel of the reference image. The output image of the switch 614 is stored in the frame memory 201 (see FIG. 2), and the output image of the switch 614 is used as a reference image in the processing of the next frame.

  The above-described embodiment of the present invention will be described in more detail based on a specific example of processing. First, still image processing, for example, processing for acquiring three still images by continuous shooting and performing noise reduction will be described with reference to the flowchart of FIG. 10 and FIG. 11. Three still images taken at each of times t, t + 1, and t + 2 are denoted as Org (t), Org (t + 1), and Org (t + 2), respectively.

  In the first step S1 in the flowchart, the motion prediction device 202 performs motion prediction using Org (t) as a reference image and Org (t + 1) as a reference image to obtain a motion vector MV. A process for obtaining a motion vector MV indicating the motion of the reference image Org (t + 1) with respect to the base image Org (t) is shown as motion prediction ST1 in FIG.

In step S2, the motion compensation device 203 performs motion compensation on the reference image Org (t + 1) using the motion vector MV, and obtains an MC image MC (t + 1). In FIG. 11, the process of obtaining the MC image is shown as step ST2.

In step S3, the addition determination device 204 performs addition determination from the reference image Org (t) and the MC image MC (t + 1) that has been aligned by motion compensation, and an addition weight is obtained. In FIG.
The process of obtaining the addition weight is shown as step ST3.

In step S4, the reference image Org (t) and the MC image MC (t + 1) are weighted and added by the addition weight in the adding device 205. In FIG. 11, the addition process is shown as step ST4. By the addition, an NR image NR (t) after noise reduction is obtained.

Next, in step S5, the motion prediction device 202 performs motion prediction using the NR image NR (t) as a reference image and the still image Org (t + 2) as a reference image, thereby obtaining a motion vector MV. The process of obtaining the motion vector MV indicating the motion of the reference image Org (t + 2) with respect to the standard image NR (t) is shown as motion prediction ST5 in FIG.

In step S6, the motion compensator 203 performs motion compensation on the reference image Org (t + 2) using the motion vector MV to obtain an MC image MC (t + 2). In FIG. 11, the process of obtaining the MC image is shown as step ST6.

In step S7, the addition determination device 204 performs addition determination from the reference image Org (t) and the MC image MC (t + 2) that has been aligned by motion compensation, and an addition weight is obtained. In FIG.
The process of obtaining the addition weight is shown as step ST7.

In step S8, the reference image Org (t) and the MC image MC (t + 2) are weighted and added by the addition weight in the adding device 205. In FIG. 11, the addition process is shown as step ST8. By the addition, an NR image NR ′ (t) after noise reduction is obtained.

  As described above, other images are aligned by motion compensation with respect to the reference image, and the reference image and the MC image are added. The addition result NR image is used as a reference image for the next processing, and an NR image is obtained by the same processing. Thus, in the case of a still image, recursive processing is performed.

  10 and 11 described above, motion prediction and motion compensation are performed in units of blocks (for example, macroblocks), and addition determination processing and addition processing are performed in units of pixels. The same applies to the moving image processing described below. The above-described example is an example in which three still images are input, but the same recursive processing is performed even if two or four or more still images are input.

Next, the moving image processing that adds two moving images will be described with reference to the flowchart of FIG. 12 and FIG. The latest image (frame) taken at time t is represented as Org (t), and the image after processing one frame before is represented as NR (t−1). In the above-described still image processing, a plurality of still images are obtained by continuous shooting, but in the moving image processing, images are continuously captured, and each frame is input at a predetermined frame rate, for example, at an interval of 1/30 seconds. .

In the first step S11, the motion prediction device 202 uses Org (t) as a reference image and NR (t-1).
As a reference picture, motion prediction is performed to obtain a motion vector MV. In FIG. 13, the process of obtaining the motion vector MV indicating the motion of the reference image NR (t-1) with respect to the reference image Org (t) is motion prediction ST1.
It is shown as 1.

  In step S12, the motion compensation device 203 performs motion compensation on the reference image NR (t-1) using the motion vector MV, and obtains an MC image MC (t-1). In FIG. 13, the process for obtaining the MC image is shown as step ST12.

In step S13, the addition determination device 204 performs addition determination from the reference image Org (t) and the MC image MC (t-1) that has been aligned by motion compensation, and an addition weight is obtained. In FIG. 13, the process of obtaining the addition weight is shown as step ST13.

In step S14, the reference image Org (t) and the MC image MC (t-1) are
Weighted addition is performed by the addition weight. In FIG. 13, the addition process is shown as step ST14. By the addition, an NR image (NR (t)) after noise reduction is obtained. The next frame image Org (t + 1) is processed in the same manner as described above using the NR image NR (t) as a reference image.

  As described above, during the moving image processing, the past image is aligned and added to the image at the latest time. In a video, since the time for performing motion prediction for the number of times of addition is small, motion prediction is performed only for the previous image, and for images at a time two or more images away, the motion prediction is added for the previous image. The MC image is stored, and motion compensation is performed with reference to the motion vector with the previous image. Thus, the number of motion predictions can be performed once for one image.

The flowchart in FIG. 14 and FIG. 15 show moving image processing, which is processing for adding three moving images. The most recent image (frame) taken at time t is represented as Org (t), the image after processing one frame before is represented as NR (t-1), added at the previous one, and stored. The rendered MC image is denoted as MC (t-2).

In the first step S21, the motion prediction apparatus 202 uses Org (t) as a reference image and NR (t-1).
As a reference picture, motion prediction is performed to obtain a motion vector MV. In FIG. 15, the process of obtaining the motion vector MV indicating the motion of the reference image NR (t-1) with respect to the reference image Org (t) is motion prediction ST2.
It is shown as 1.

In step S22, the motion compensation device 203 performs motion compensation on the reference image NR (t-1) using the motion vector MV, and obtains an MC image MC (t-1). In FIG. 15, the process of obtaining the MC image MC (t-1) is shown as step ST22.

In step S23, the addition determination device 204 performs addition determination from the reference image Org (t) and the MC image MC (t-1) that has been aligned by motion compensation, and an addition weight is obtained. In FIG. 15, the process of obtaining the addition weight is shown as step ST23.

In step S24, the reference image Org (t) and the MC image MC (t-1) are
Weighted addition is performed by the addition weight. In FIG. 15, addition processing is shown as step ST24. By the addition, an NR image (NR (t)) after noise reduction is obtained.

In step S25, the motion compensation device 203 performs motion compensation on the MC image MC (t-2) using the motion vector MV, and obtains an MC image MC ′ (t-2). In FIG. 15, the process of obtaining the MC image MC ′ (t−2) is shown as step ST25.

In step S26, the addition determination device 204 performs addition determination from the NR image NR (t) as the reference image and the MC image MC ′ (t−2), and an addition weight is obtained. In FIG. 15, the process of obtaining the addition weight is shown as step ST26.

In step S27, the NR image NR (t) and the MC image MC ′ (t−2) are added by the adding device 205.
Weighted addition is performed by the addition weight. In FIG. 15, addition processing is shown as step ST27. By the addition, an NR image NR ′ (t) after noise reduction is obtained.

  Like the processing described above, motion prediction is performed only once in step S21 (ST21), and motion compensation is performed by applying the motion vector MV to NR (t-1) and MC (t-2). Three images can be added without greatly increasing the amount.

The flowchart of FIG. 16 and FIG. 17 show the process of adding three moving images and the hidden area process. The most recent image (frame) taken at time t is represented as Org (t), the image after processing one frame before is represented as NR (t-1), added at the previous one, and stored. The displayed MC picture is denoted as MC ′ (t-2).

  The motion prediction in step S31 (ST31) and the motion compensation in step S32 (ST32) are processes corresponding to steps S21 (ST21) and S22 (ST22) in FIGS.

  The addition determination in step S33 (ST33) and the addition in step S34 (ST34) are processes corresponding to steps S23 (ST23) and S24 (ST24) in FIGS. By the addition, an NR image NR (t) after noise reduction is obtained.

In step S35, the hidden region processing unit 206 performs hidden region processing on the MC image MC (t-1) for the noise reduction processing of the next frame, and the MC image MC '(t-1) is obtained. . FIG.
7, the range of the hidden area hidden by the moving object by the motion compensation is indicated by R, and the hidden area processing is indicated as step ST35.

  In step S36, the motion compensation device 203 performs motion compensation on the MC image MC ′ (t−2) using the motion vector MV to obtain an MC image MC ″ (t−2). In FIG. 17, the process of obtaining the MC image MC ″ (t−2) is shown as step ST36.

  In step S37, the addition determination device 204 performs addition determination from the NR image NR (t) as the reference image and the MC image MC ″ (t−2), and an addition weight is obtained. In FIG. 17, the process of obtaining the addition weight is shown as step ST37.

  In step S38, the NR image NR (t) and the MC image MC ″ (t−2) are weighted and added by the addition weight in the adding device 205. In FIG. 17, the addition process is shown as step ST38. By the addition, an NR image NR ′ (t) after noise reduction is obtained.

As in the processing described above, hidden region processing is performed on the image MC (t-1) obtained by motion compensation of the processing image NR (t-1) before the unit time (one frame), and MC '(t-1) is calculated. By creating the image and using it in the noise reduction process for the next time, it is possible to add three images while suppressing the propagation of image deterioration in the hidden area.

  Next, another embodiment of the present invention will be described. In the above-described embodiment of the present invention, an MC image in which motion compensation is performed for other images with respect to a reference image in units of blocks is generated, and addition determination in units of pixels is performed, so that camera shake and moving subjects can be detected. In contrast, a noise reduction effect can be obtained. However, under certain conditions, there is a problem that the deterioration due to motion estimation and motion compensation is noticeable.

  That is, in the case of adding a large number of MC images, when almost the same block distortion occurs at the same position of a plurality of MC images and the pixel value of the distortion is close to the pixel value of the reference image, the addition is performed. Deterioration accumulates and block-like distortion occurs in the processed image. Actually, when multiple images are shot in a situation where there is little camera movement, there is a moving subject only in the reference image, and in other images the moving subject is not shot. Cannot be estimated correctly, and almost the same block distortion occurs in almost all MC images.

  Furthermore, under certain conditions, there may be a problem that the conspicuous noise reduction effect is insufficient. That is, the spatially large color noise generated in the flat portion has a large amplitude in addition to the magnitude, and cannot be determined to increase the addition weight in the addition determination, and noise removal is difficult.

  In order to solve these problems, the following operation is performed when the addition determination of the reference image and the MC image is performed in units of pixels. First, considering the block distortion generated in the MC image, the block boundary is not noticeable due to the noise in the area where the noise exists, and the block boundary is conspicuous in the area where the noise is low. In view of this, a flat area with little noise, that is, an area where noise is conspicuous is detected, and correction is performed to make the addition weight smaller in that area. By correcting the addition weight, block distortion occurring in the MC image can be suppressed.

  A process of reducing spatially large color noise generated in the flat portion will be described. When the color difference difference between the reference image and the MC image is large, the area that is flat and the average value of the color difference difference calculated over the area where color noise occurs is determined as color noise, and the addition weight is set. Increase.

  FIG. 18 shows the configuration of a noise reduction apparatus according to another embodiment of the present invention that performs processing for correcting the addition weight in this way. The noise reduction apparatus includes a frame memory 201, a motion prediction apparatus 202, a motion compensation apparatus 203, an addition determination apparatus 204, an addition weight correction apparatus 207, and an addition apparatus 205. Note that a hidden area processing device may be provided as in the embodiment.

  The configuration other than the addition weight correction device 207 that corrects the addition weight output from the addition determination device 204 is the same as that of the above-described embodiment. That is, an addition weight correction apparatus that corrects the addition weight output from the addition determination apparatus (configuration shown in FIG. 3) is shown. The configuration of the imaging device (FIG. 1), the configuration of the noise reduction device (FIG. 2), and the configuration and characteristics related to the addition determination device (FIG. 4 and the like) in the above-described embodiment are the same as in other embodiments. It is the same.

  FIG. 19 shows an addition weight correction device that corrects the addition weight output from the addition determination device (configuration shown in FIG. 3). The addition weight output from the addition determination device is supplied to the multiplier 332. The addition weight generated by the addition determination unit 331 in consideration of the suppression of the color noise described above is supplied to the multiplier 332.

  An output signal of the multiplier 332 is supplied to the clipping circuit 333, and a value exceeding 1 is clipped to 1 by the clipping circuit 333. An output signal of the clip circuit 333 is supplied to the multiplier 334. To the multiplier 334, the addition weight generated by the addition determination unit 336 taking into account the suppression of the MC image block noise described above is supplied. An addition weight corrected for noise suppression is obtained at the output of the multiplier 334. This addition weight is supplied to the addition device 205.

  The output signal of the adder 335 is supplied to the addition determiner 336 for countermeasure against block noise. As one input of the adder 335, the moving variance obtained by the luminance moving variance computing unit 320 with respect to the luminance reference image via the prefilter 300 is supplied, and as the other input, the output signal of the upsampling circuit 338 is supplied. The

  The output signal of the adder 337 is supplied to the upsampling circuit 338. As one input of the adder 337, the moving variance obtained by the color difference moving variance computing unit 420 with respect to the color difference (Cb) reference image via the prefilter 400 is supplied, and as the other input, the color difference via the prefilter 400 is supplied. (Cr) The moving dispersion obtained by the color difference moving dispersion calculator 520 with respect to the reference image is supplied. The size of the rectangular area for calculating the moving variance of luminance and chrominance is set to be the same. The upsampling circuit 338 is provided for interpolation with respect to a color difference signal whose number of pixels is smaller than that of a luminance signal.

  The addition determiner 336 corrects the addition weight as shown in FIG. 20 for the input signal (sum of variance values) supplied from the adder 335. When the input signal is smaller than the determination threshold TH, an addition weight that increases corresponding to the input signal is output. When the input signal is greater than or equal to the threshold TH, an addition weight of 1 is output. The addition weight set to 1 has no effect even if it is multiplied by the addition weight from the addition determination device.

  The region where the sum of the variance values of the reference image is small can be estimated as a flat region with little noise, that is, a region where noise is easily noticeable. In such a region, block distortion occurring in the MC image can be suppressed by performing correction to make the addition weight smaller. The addition weight set to be smaller than 1 is multiplied by the addition weight from the addition determination device to reduce the addition weight.

  The luminance reference image variance from the luminance moving variance computing unit 320 and the moving average from the color difference moving average computing unit 340 are input to the addition determination unit 331 for measures against color noise. The pixel absolute value difference of the color difference (Cb) reference image and the pixel absolute value difference of the color difference (Cr) reference image are added by the adder 339, and the moving average of the color difference difference absolute values output from the adder 339 is the moving average. Calculation is performed in the calculator 340. Moving average calculators 320 and 340 obtain a moving average of a rectangular area centered on a pixel whose value is to be calculated. The size of the rectangular area is larger than the color noise to be removed.

  The addition determiner 331 has a hardware or software configuration for performing processing as shown in FIG. First, in the first step S41, it is determined whether or not the luminance dispersion value is within a certain threshold range. Being within the threshold range means a flat region. If it is not a flat region, the addition weight is set to 1 in step S44, and the process ends. The addition weight set to 1 has no effect even if it is multiplied by the addition weight from the addition determination device.

  If it is determined in step S41 that the region is a flat region, it is determined whether or not the moving average of the color difference difference absolute value is within a certain threshold range. If the determination result of step S41 is affirmative, it is determined as a color noise region, and the process of step S43 is performed. If the determination result of step S41 is negative, the process of step S44 (addition weight = 1) is performed.

  In step S43, a value greater than 1 is set as the addition weight, and the process ends. The addition weight set to be greater than 1 is multiplied by the addition weight from the addition determination device, increasing the addition weight and increasing the color noise removal effect. The specific numerical values of the threshold and the addition weight to be set are set in consideration of the visual influence. If the result of multiplying an addition weight greater than 1 exceeds 1, the value is set to 1 by the clip circuit 333.

  According to the other embodiment of the present invention described above, in addition to the noise reduction effect similar to the one embodiment, the occurrence of block distortion can be suppressed, and the spatially large size generated in the flat portion can be suppressed. Color noise can be reduced.

  Although the embodiment of the present invention has been specifically described above, the present invention is not limited to the above-described embodiment, and various modifications based on the technical idea of the present invention are possible. For example, in the example described above, the reference image and the MC image are weighted averaged, but may be added. In other words, a dark image with no camera shake or noise may be positioned and added for determination, and a plurality of images may be added according to the addition weight, and finally normalized for each pixel.

It is a block diagram which shows the whole structure of one embodiment of this invention. It is a block diagram of an example of the noise reduction apparatus in one embodiment of this invention. It is a block diagram of an example of the addition determination apparatus in one embodiment of this invention. It is a basic diagram which shows notionally the table of the pixel value versus noise amount with which the noise amount estimator in an addition determination apparatus is provided. It is a block diagram which shows an example of the pixel addition determination device in an addition determination apparatus. It is a basic diagram which shows some examples of the determination curve which can be used. It is an approximate line figure showing an example of a judgment curve. It is a block diagram which shows an example of the addition apparatus in an addition determination apparatus. It is a block diagram which shows an example of the hidden region processing apparatus in an addition determination apparatus. It is a flowchart for demonstrating the still image process in one embodiment of this invention. It is a basic diagram for demonstrating the reference relationship in a still image process. It is a flowchart for demonstrating the moving image process which adds two moving images in one embodiment of this invention. It is a basic diagram for demonstrating the reference relationship in the moving image process which adds two moving images. It is a flowchart for demonstrating the moving image process which adds three moving images in one embodiment of this invention. It is a basic diagram for demonstrating the reference relationship in the moving image process which adds three moving images. It is a flowchart for demonstrating the moving image process (including a hidden area process) which adds the three moving images in one embodiment of this invention. It is a basic diagram for demonstrating the reference relationship in the moving image process (including a hidden area process) which adds three moving images. It is a block diagram of an example of the noise reduction apparatus in one embodiment of this invention. It is a block diagram which shows the structural example of the addition weight correction | amendment apparatus of other embodiment of this invention. It is a basic diagram for demonstrating the block noise reduction process in other embodiment of this invention. It is a basic diagram for demonstrating the color noise reduction process in other embodiment of this invention.

DESCRIPTION OF SYMBOLS 101 ... Imaging device 102 ... Camera signal processing part 103 ... Noise reduction apparatus 104 ... Camera control part 201 ... Frame memory 202 ... Motion estimation apparatus 203 ... Motion compensation apparatus 204- ..Addition determination device 205... Addition device 206... Hidden region processing device 207... Addition weight correction device 300, 301, 400, 401, 500, 501 ... Prefilter 310, 311. Moving average computing unit 410, 411, 510, 511 ... chrominance moving dispersion computing unit 320, 321 ... luminance moving dispersion computing unit 420, 421, 520, 521 ... chrominance moving dispersion computing unit 313, 323, 413 , 423, 513, 523 ... absolute value calculator 305 ... luminance noise amount estimator 405 ... color difference (Cb) noise amount estimator 5 05: Color difference (Cr) noise amount estimator 304 ... Luminance pixel addition decision unit 314 ... Luminance moving average addition decision unit 324 ... Luminance movement variance addition decision unit 404, 504 ... Color difference pixel addition Determinator 414, 514 ... Color difference moving average addition determiner 424, 524 ... Color difference moving dispersion addition adder 338, 527 ... Upsampling circuit 601 ... Noise amount-> decision threshold converter 602 ... Determination curve generator 331... Addition determination device for color noise countermeasure 336... Addition determination device for MC image block distortion countermeasure

Claims (11)

An image processing method in which an image with reduced noise is output using first and second still images that are temporally continuous,
Performing block-based motion compensation on the second still image;
Generating a table for setting an addition weight for a pixel difference absolute value between the first still image and the second still image subjected to the motion compensation, and setting the addition weight using the table ;
The second still image subjected to the first still image and the motion compensation, possess and obtaining an image after addition or weighted average processing according to the weighting factor,
An image processing method in which a relationship between noise amounts corresponding to pixel values under a plurality of shooting conditions is stored in advance, and the addition weight is set by defining the table according to a current shooting condition . An image processing method for generating a noise-reduced image using first, second, and third still images that are temporally continuous,
Performing block-based motion compensation on the second still image;
A table for setting a first addition weight is generated for an absolute value of a pixel difference between the first still image and the second still image subjected to the motion compensation, and the first addition weight is set using the table. And steps to
Adding the first still image and the second still image subjected to motion compensation according to the first addition weight to obtain a processed image; and for the third still image, a block unit Performing motion compensation of
A table for setting a second addition weight is generated for a pixel difference absolute value between the first still image and the third still image subjected to the motion compensation, and a second addition weight is set using the table. And steps to
Said third still picture that has been subjected to the predetermined image and the motion compensation after the treatment, possess and obtaining an image after addition or weighted average processing according to the second weighting factor,
The relationship of the noise amount corresponding to the pixel value under each shooting condition of a plurality of shooting conditions is stored in advance, and the first addition weight and the second addition weight are defined by defining the table according to the current shooting condition. Image processing method to set . An image processing method for reducing noise for moving images input at unit time intervals,
Performing block-unit motion compensation on an image after processing one unit time ago;
A step of generating a table for setting an addition weight for a pixel difference absolute value between the current moving image and the processed image of the unit time before the motion compensation, and setting the addition weight using the table When,
Adding the current image and the processed image one unit time before performing the motion compensation according to the addition weight, or obtaining a processed image corresponding to the current moving image. Yes, and
An image processing method in which a relationship between noise amounts corresponding to pixel values under a plurality of shooting conditions is stored in advance, and the addition weight is set by defining the table according to a current shooting condition . An image processing method for reducing noise for moving images input at unit time intervals,
A motion vector in block units is obtained from the current moving image and a processed image one unit time before, and the motion is calculated with respect to the processed image one unit time before and the motion compensated image two units time ago. Performing motion compensation with vectors;
A table for setting a first addition weight is generated for a pixel difference absolute value between the current moving image and the image after processing one unit time before the motion compensation is performed, and a first addition is performed using the table. Setting weights;
Adding the current moving image and the processed image of one unit time before performing motion compensation in accordance with the first addition weight to obtain an image after the first processing;
Generating a table for setting a second addition weight with respect to an absolute value of pixel difference between the motion-compensated image after motion compensation two units before and the image after the first processing; And setting a second addition weight,
The motion-compensated image after motion compensation two units before and the image after the first processing are added or weighted averaged according to the second addition weight, and the second corresponding to the current moving image. Obtaining the processed image of
The relationship of the noise amount corresponding to the pixel value under each shooting condition of a plurality of shooting conditions is stored in advance, and the first addition weight and the second addition weight are defined by defining the table according to the current shooting condition. Image processing method to set . And a correction step of correcting the addition weight or the first and second addition weights,
The correction step includes
Detecting when block distortion occurs in the processed image,
The image processing method according to claim 1, wherein the addition weight or the first and second addition weights are corrected so as to reduce the detected block distortion. And a correction step of correcting the addition weight or the first and second addition weights,
The correction step includes
Detecting color noise of a large size generated in a flat portion in the image after the processing;
The image processing method according to claim 1, wherein the addition weight or the first and second addition weights are corrected so as to reduce the detected color noise. Estimating the amount of noise by having a plurality of pixel value-to-noise amount table showing the amount of noise relationship corresponding to the pixel values, selecting the pixel value-to-noise amount table according to the imaging condition,
Generating a table indicating a relationship between the pixel difference absolute value and the addition weight regarding two pixels in two spatially identical positions of two images according to the estimated noise amount;
Obtaining an addition weight for the index value by the table;
The image processing method according to any one of claims 1 to 4 having a. An image processing apparatus that generates a noise-reduced image using first and second still images that are temporally continuous,
A motion compensation device that performs motion compensation on a block basis for the second still image;
A determination device configured to generate a table for setting an addition weight for an absolute value of a pixel difference between the first still image and the second still image subjected to the motion compensation, and to set the addition weight using the table ; ,
The second still image subjected to the first still image and the motion compensation, possess an adding device for obtaining an image after addition or weighted average processing according to the weighting factor,
An image processing apparatus in which a relationship between noise amounts corresponding to pixel values under a plurality of shooting conditions is stored in advance, and the addition weight is set by defining the table according to a current shooting condition . An image processing apparatus that generates a noise-reduced image using first, second, and third still images that are temporally continuous,
A motion compensation device, a determination device, and an addition device;
The motion compensation device performs motion compensation on a block basis for the second still image,
A table for setting a first addition weight for a pixel difference absolute value between the first still image and the second still image subjected to motion compensation is generated by the determination device, and a first setting is made using the table . Set the addition weight,
The addition device performs addition or weighted averaging on the first still image and the second still image subjected to motion compensation according to the first addition weight to obtain a processed image,
The motion compensation device performs motion compensation on a block basis for the third still image,
The determination device generates a table for setting a second addition weight with respect to a pixel difference absolute value between the first still image, the processed image, and a third still image subjected to motion compensation, and Use to set the second additive weight,
By the addition device, the processed image and the third still image subjected to motion compensation are subjected to addition or weighted average according to the second addition weight to obtain a processed image ,
The relationship of the noise amount corresponding to the pixel value under each shooting condition of a plurality of shooting conditions is stored in advance, and the first addition weight and the second addition weight are defined by defining the table according to the current shooting condition. An image processing apparatus configured to set . An image processing apparatus that performs noise reduction on a moving image input at unit time intervals,
A motion compensation device, a determination device, and an addition device;
The motion compensation device generates a table for setting an addition weight for the pixel difference absolute value of the image after processing one unit time before, and performs motion compensation on a block basis for the table .
The determination device generates a table for setting an addition weight for the absolute value of the pixel difference between the current moving image and the image after the processing performed one unit time before motion compensation is performed, and the addition is performed using the table. Set weights,
The added image is subjected to addition or weighted averaging of the current moving image and the processed image before one unit time before performing motion compensation, and the processed image corresponding to the current moving image. the resulting,
Image processing in which the relationship of the amount of noise corresponding to the pixel value under each shooting condition of a plurality of shooting conditions is held in advance, and the addition weight is set by defining the table according to the current shooting condition apparatus. An image processing apparatus that performs noise reduction on a moving image input at unit time intervals,
A motion compensation device, a determination device, and an addition device;
A motion vector in block units is obtained from the current moving image and a processed image one unit time before, and the motion is calculated with respect to the processed image one unit time before and the motion compensated image two units time ago. Motion compensation by vector,
A table for setting a first addition weight for a pixel difference absolute value between the current moving image and the image after processing one unit time before performing motion compensation is generated, and the table is used as a shooting condition. Set a first additive weight using a noise amount based on
The current image and the processed image before the unit time before performing motion compensation are added or weighted average according to the first addition weight to obtain an image after the first processing,
Generating a table for setting a second addition weight with respect to an absolute value of pixel difference between the motion-compensated image after motion compensation two units before and the image after the first processing; To set the second addition weight,
The motion-compensated image after motion compensation two units before and the image after the first processing are added or weighted averaged according to the second addition weight, and the second corresponding to the current moving image. obtain an image of the post-processing of,
The relationship of the noise amount corresponding to the pixel value under each shooting condition of a plurality of shooting conditions is stored in advance, and the first addition weight and the second addition weight are defined by defining the table according to the current shooting condition. An image processing apparatus configured to set .

Images