WO2014054273A1 - 画像ノイズ除去装置、および画像ノイズ除去方法 - Google Patents
画像ノイズ除去装置、および画像ノイズ除去方法 Download PDFInfo
- Publication number
- WO2014054273A1 WO2014054273A1 PCT/JP2013/005844 JP2013005844W WO2014054273A1 WO 2014054273 A1 WO2014054273 A1 WO 2014054273A1 JP 2013005844 W JP2013005844 W JP 2013005844W WO 2014054273 A1 WO2014054273 A1 WO 2014054273A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- noise
- reliability
- noise removal
- spatial
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 65
- 239000000203 mixture Substances 0.000 claims abstract description 58
- 238000012545 processing Methods 0.000 claims description 64
- 238000004364 calculation method Methods 0.000 claims description 63
- 238000002156 mixing Methods 0.000 claims description 34
- 239000013598 vector Substances 0.000 claims description 23
- 230000002123 temporal effect Effects 0.000 description 57
- 238000001514 detection method Methods 0.000 description 16
- 238000010586 diagram Methods 0.000 description 11
- 230000000694 effects Effects 0.000 description 10
- 238000009499 grossing Methods 0.000 description 7
- 239000000470 constituent Substances 0.000 description 5
- 206010047571 Visual impairment Diseases 0.000 description 4
- 238000004590 computer program Methods 0.000 description 4
- 238000007796 conventional method Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N5/00—Details of television systems
- H04N5/14—Picture signal circuitry for video frequency region
- H04N5/21—Circuitry for suppressing or minimising disturbance, e.g. moiré or halo
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20172—Image enhancement details
- G06T2207/20182—Noise reduction or smoothing in the temporal domain; Spatio-temporal filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Definitions
- the present invention relates to an image noise removing apparatus and an image noise removing method.
- Filter processing for removing noise from image data includes a temporal noise removal filter and a spatial noise removal filter.
- a method of switching between these noise removal filters Patent Document 1 discloses a method of selecting one of the processing results of the temporal noise removal filter and the spatial noise removal filter using the result of motion detection.
- Patent Document 2 discloses a method of outputting the final processing result after determining whether to use the processing result of the temporal noise removal filter using the processing result of the spatial noise removal filter.
- noise When noise is removed by the temporal noise removal filter, there is a problem that noise (for example, afterimage or moving image discomfort) may be newly generated. In other words, there is a problem that it is difficult to perform temporal noise removal processing with an intensity suitable for removing noise on an image.
- the present invention provides an image noise removing apparatus that applies time noise removal processing with an appropriate intensity to an image.
- An image noise removing apparatus is an image noise removing apparatus for removing noise included in a second image after removing noise from the first image, and includes pixels included in the second image.
- a spatial noise removing unit that generates a spatial noise-removed image by performing an operation for removing noise included in the second image using a value, the spatial noise-removed image, the second image, and the first image Based on the first noise-removed image from which noise of one image has been removed, a reliability calculation unit that calculates the reliability of movement in the second image, and the second image and the first noise-removed image
- a time blending unit that removes noise included in the second image by performing weighted addition processing based on the reliability.
- the image noise removal apparatus and method of the present invention it is possible to perform noise removal that optimally applies the time noise removal filter to the target image by adaptively changing the filter strength of the time noise removal filter.
- FIG. 1 is a configuration diagram of an image noise removing apparatus according to the first embodiment.
- FIG. 2 is a configuration diagram of the reliability calculation unit in the first embodiment.
- FIG. 3 is an explanatory diagram of a relationship between a pixel value and a vector in the first embodiment.
- FIG. 4 is a diagram showing a probability distribution of noise.
- FIG. 5 is an example of the correspondence between ⁇ tnr and P in the first embodiment.
- FIG. 6 is a flowchart showing the flow of processing in the first embodiment.
- FIG. 7 is a configuration diagram of the image noise removing apparatus according to the second embodiment.
- FIG. 8 is a flowchart showing the flow of processing in the second embodiment.
- FIG. 9 is an example of a configuration diagram of an image noise removing device in each embodiment.
- Image (moving image, still image) data includes noise (noise) depending on the shooting environment or the characteristics of the image sensor.
- Types of noise include light shot noise, dark noise, fixed pattern noise, or circuit noise. Since the quality (image quality) of the image deteriorates due to this noise, noise removal for removing or reducing the noise of the image data is performed by the filter processing.
- noise removal filters are spatial noise removal filters that are processed using data in one image.
- a temporal noise removal filter that reduces or removes noise by a smoothing process using correlation between the frames is also used.
- the temporal noise removal method there is a method of performing weighted averaging of pixel data of interest and pixel data of a frame existing in time.
- a temporal noise removal filter is selectively used if the pixel data of interest is a stationary part, and a spatial noise removal filter is selectively used if it is a moving part.
- Patent Document 1 discloses a method of selecting one of processing results of a temporal noise removal filter and a spatial noise removal filter using a motion detection result. Disclosure.
- Patent Document 2 discloses a method of outputting the final processing result after determining whether to use the processing result of the temporal noise removal filter using the processing result of the spatial noise removal filter.
- the present invention provides an image noise removing apparatus that applies time noise removal processing with an appropriate intensity to an image.
- an image noise removing apparatus for removing noise included in a second image after removing noise from the first image.
- a spatial noise removing unit that generates a spatial noise-removed image by performing an operation for removing noise included in the second image using a pixel value included in the second image; and the spatial noise-removed image
- a reliability calculation unit that calculates a reliability of movement in the second image based on the second image and the first noise-removed image from which noise of the first image has been removed; and the second image
- a time blending unit that removes noise included in the second image by performing weighted addition processing based on the reliability with respect to the first noise-removed image.
- the movement in the second image is detected with high accuracy using the spatial noise-removed image obtained by spatially removing the noise of the second image, Depending on the detection result, temporal noise removal can be performed on the second image. Since the accuracy of motion detection in the second image is high, the accuracy of temporal noise removal can be improved. Therefore, the image noise removal device can perform temporal noise removal processing with appropriate intensity on the image.
- the detection of motion in the second image according to the conventional technique is performed by using the second image and the first noise-removed image, but it is known that the detection result is inaccurate.
- the accuracy of the motion detection result can be improved by using the spatial noise-removed image of the second image in addition to the above two images.
- the reliability calculation unit calculates a larger value as the reliability as the movement of the pixels included in the second image is smaller, and the time blending unit calculates the first as the reliability is higher.
- a weighted addition process in which the weight of the noise-removed image is increased is performed as the weighted addition process.
- the motion in the second image is smaller, it is possible to perform the temporal noise removal process in which the weight of the first noise removed image is increased.
- the noise of the image after the temporal noise removal process can be reduced by increasing the weight of the first noise-removed image from which noise has already been removed.
- the reliability calculation unit calculates a larger value as the reliability as the difference between pixel values at the same position included in each of the first noise-removed image and the second image is smaller.
- the reliability calculation unit calculates a larger value as the reliability as the difference between pixel values at the same position included in each of the first noise-removed image and the spatial noise-removed image is smaller.
- the image noise removing device increases the weight of the first noise-removed image in the temporal noise removal when it is determined that the possibility that there is a motion in the second image is smaller. Since the noise has already been removed from the first noise-removed image, the noise of the image after the temporal noise removal processing can be reduced.
- the reliability calculation unit calculates the reliability based on a ratio of the difference to a deviation of a pixel value distribution that is a predetermined deviation and changes due to noise.
- the image noise removing device can determine the size by using the ratio to the deviation of the distribution of the pixel value that changes due to the noise when determining the difference of the pixel value.
- the pixel value includes a plurality of components
- the reliability calculation unit includes, as a component, a difference between pixel values of pixels at the same position in each of the spatial noise-removed image and the second image. The smaller the angle formed between the vector and the second vector having as a component the pixel value difference of the pixel at the same position in each of the first noise-removed image and the second image, the larger the value as the reliability. calculate.
- the image noise removal device includes a first vector corresponding to a difference between the spatial noise removal image and the second image, and a second vector corresponding to the difference between the first noise removal image and the second image. It is determined that the possibility that there is a motion in the second image is smaller as the formed angle is smaller, and the weight of the first noise-removed image in the temporal noise removal is increased. Since the noise has already been removed from the first noise-removed image, the noise of the image after the temporal noise removal processing can be reduced.
- the first noise-removed image is an image obtained by removing noise included in the first image by the image noise removing device.
- the image noise removal device can use an image obtained by performing temporal noise removal processing on the first image by the image noise removal device as the first noise removal image.
- the first image is one of images constituting a moving image
- the second image is one of images constituting the moving image
- the images are arranged in time order. It is an image immediately after the first image.
- a temporal noise removal process is performed on successive images in consideration of a noise removal result of an image immediately before the image. Can be applied.
- the image noise removal device further includes a spatial blend unit that generates a spatial blend image by performing weighted addition processing based on the reliability on the second image and the spatial noise removal image.
- the time blending unit further performs weighted addition processing based on the reliability on the spatial blended image and the first noise-removed image from which noise included in the first image has been removed. , Noise included in the second image is removed.
- the reliability calculation unit calculates a larger value as the reliability as the movement of a pixel included in the second image is smaller, and the spatial blending unit calculates the second as the reliability is higher.
- a weighted addition process in which the weight of the image is increased is performed as the weighted addition process, and the time blending unit performs the weighted addition process in which the weight of the first noise-removed image is increased as the reliability increases. This is performed as a weighted addition process.
- An image noise removal method is an image noise removal method for removing noise included in a second image after noise removal of the first image, and is included in the second image.
- a time blending step for removing noise included in the second image by performing weighted addition processing based on the reliability.
- a program according to an aspect of the present invention is a program for causing a computer to execute the image noise removal method described above.
- An integrated circuit is an integrated circuit for removing noise included in the second image after removing noise from the first image, and the pixel value included in the second image is determined.
- a spatial noise removal unit that generates a spatial noise-removed image by performing an operation for removing noise included in the second image, the spatial noise-removed image, the second image, and the first image.
- a reliability calculation unit that calculates the reliability of the movement in the second image, and the second image and the first noise-removed image,
- FIG. 1 is a configuration diagram of an image noise removing apparatus according to the present embodiment.
- the image noise removing apparatus 1 includes an input image terminal 10, a spatial noise removing unit 20, a reliability calculating unit 30, a time blending rate calculating unit 40, a time blending unit 50, and a buffer. 60 and an output image terminal 70.
- the input image terminal 10 receives an image (hereinafter referred to as “input image”) input to the image noise removal device 1. Specifically, the input image terminal 10 receives the pixel value (pixel data) of the target pixel and the pixel values (pixel data) of the surrounding pixels. Note that the input image corresponds to the second image.
- the spatial noise removal unit 20 performs noise removal by smoothing using the pixel value of the target pixel received by the input image terminal 10 and the pixel values of the surrounding pixels.
- the image from which noise has been removed in this way corresponds to a spatial noise-removed image.
- the reliability calculation unit 30 calculates the probability that the target pixel received by the input image terminal 10 is stationary with respect to the pixel at the same position in the past input image (corresponding to the first image). That is, the reliability calculation unit 30 receives the pixel value of the target pixel received by the input image terminal 10, the pixel value at the same position in the image obtained by removing noise from the past image recorded in the buffer, and the space. The pixel value at the same position in the image after the noise removal unit 20 performs noise removal is used as an input, and using these, the target pixel is stationary relative to the pixel at the same position in the noise removal signal for the past image. Calculate the probability of being. The calculation method will be described later.
- the pixel is stationary means that the photographing object corresponding to the pixel does not change. That is, “the pixel is stationary” when the photographing objects corresponding to the pixels in each of the two images are the same. Furthermore, it can be said that “the pixel is stationary” even when the pixel values are different because noises of different magnitudes are added to these two pixels. On the other hand, when the photographing objects corresponding to the pixels in the two images are different, it is said that “the pixels are not stationary”.
- one image pixel is stationary relative to the other image pixel. It can be said. Furthermore, when different noise is added to the pixel of the one image and the pixel of the other image, the pixel value is different, but in this case also, the pixel of one image is different from that of the other image. It can be said that it is stationary with respect to the pixel. On the other hand, if two images are extracted from a moving image of a subject that is actually moving (for example, a moving car or a person waving), the pixel of the one image is in the moving portion. It cannot be said that it is stationary with respect to the other image pixel.
- the “past image recorded in the buffer” means an image that has been past the input image received by the input image terminal 10 and has been subjected to noise removal processing. That is, the past image may be an image input before the input image and subjected to noise removal processing, or may be input a predetermined number of times before the input image and subjected to noise removal processing. It may be an image. At this time, the predetermined number may be the number of images considered to have a relatively small difference from the input image.
- the “past image recorded in the buffer” corresponds to the first noise-removed image.
- the time blend rate calculation unit 40 is recorded in the buffer with the pixel value received by the input image terminal 10 used by the time blend unit 50.
- a blend ratio (mixing ratio) with a pixel value of a past image is calculated.
- the time blend unit 50 receives the pixel value received by the input image terminal 10 and the pixel value at the same position in the noise removal signal for the past image recorded in the buffer.
- the output image is generated by blending with and the generated output image is output.
- FIG. 2 is a configuration diagram of the reliability calculation unit 30 in the present embodiment.
- the reliability calculation unit 30 includes a difference calculation unit 31a, a difference calculation unit 31b, a difference calculation unit 31c, and a probability calculation unit 32. Thereafter, the pixel value received by the input image terminal 10 input to the reliability calculation unit 30 is CUR, and the pixel value after the spatial noise removal unit 20 performs noise removal (spatial noise removal) is SNR, and is recorded in the buffer 60.
- the image data of the processed image will be described as PRE.
- a pixel value is composed of a plurality of color component data such as RGB or YCbCr
- a, b, and c calculated by the difference calculation units 31a, 31b, and 31c are regarded as vectors having a plurality of components. be able to.
- the vectors a, b, and c are expressed as (Equation 1).
- the subscripts y, cb, and cr indicate the Y component, Cb component, and Cr component of the pixel value, respectively.
- the vectors a, b, and c are three-dimensional vectors when the pixel value is composed of three components (for example, YCbCr or RGB format). However, for example, when there is only information about the Y component as in gray scale (when Cb and Cr are always 0), it may be treated as one-dimensional, or Cb and Cr are regarded as zero and regarded as three-dimensional. May be handled.
- FIG. 3 is an explanatory diagram of the relationship between the pixel value and the vector in the first embodiment.
- FIG. 3 shows the relationship between the pixel values CUR, SNR and PRE, and the vectors a, b and c.
- each of the pixel values CUR, SNR, and PRE is arranged at coordinates corresponding to the pixel value.
- the vector a is drawn as a vector having a CUR as a start point and an SNR as an end point.
- the vector b is drawn as a vector starting from CUR and ending at PRE.
- the vector c is drawn as a vector having the SNR as the start point and the PRE as the end point.
- the probability calculation unit 32 calculates the probability P that the pixel data is stationary according to (Equation 2).
- ⁇ , ⁇ and ⁇ are constants given from the outside. Further, ⁇ 0, ⁇ 1, and ⁇ 2 are calculated by (Expression 3), (Expression 4), and (Expression 5).
- ⁇ is a deviation of a noise model normal distribution (a distribution of pixel values changed by noise) included in the image, and is given from the outside.
- FIG. 4 shows an example of a probability distribution of noise included in an image. This probability distribution is modeled as a normal distribution of deviation ⁇ centered on the true pixel value S that does not contain noise.
- the method using only b is equivalent to using only the first factor on the right side of (Expression 2).
- a deviation from the ideal case is considered in which both the pixel value after temporal noise removal and the pixel value after spatial noise removal are correct.
- the pixel value after temporal noise removal when it is assumed that noise has been completely removed by temporal noise removal and the pixel after spatial noise removal when it is assumed that noise has been completely removed by spatial noise removal The value is considered equal. That is, in an ideal case, the difference
- the probability P is calculated by (Expression 2) by further using the above
- the above relationship does not hold, it indicates that the accuracy of time noise removal is low. That is, there is a high possibility that the pixel of interest has moved from the past image (not stationary).
- the time blend rate calculation unit 40 calculates the blend rate ⁇ tnr used in the time blend unit 50 using the probability P that the pixel value calculated by the reliability calculation unit 30 is stationary. Although ⁇ tnr is normalized by 1 and output, normalization is not necessarily required in relation to the calculation in the subsequent time blending unit 50. ⁇ tnr may be obtained by calculation using the probability P, or may be calculated using a table or correspondence prepared in advance.
- FIG. 5 is an example of the correspondence between ⁇ tnr and P in the first embodiment.
- (A) of FIG. 5 is a first example of the correspondence between ⁇ tnr and P. It shows that ⁇ tnr changes in proportion to P when P changes from 0 to 1.
- the range in which ⁇ tnr changes is, for example, 0 to 0.9.
- FIG. 5B is a second example of the correspondence between ⁇ tnr and P.
- ⁇ tnr changes in proportion to P in the interval between P and 0 to a predetermined number
- ⁇ tnr takes a constant value in the interval between P and the predetermined number.
- the range in which ⁇ tnr changes is, for example, from 0 to 0.9 as described above.
- ⁇ tnr can be changed from 0 to a predetermined number.
- the time blend rate calculation unit 40 can be omitted, and the configuration can be simplified.
- noise removal can be performed by optimally applying the temporal noise removal filter to the target image by adaptively changing the filter strength of the temporal noise removal filter.
- FIG. 6 is a flowchart showing the flow of processing in the present embodiment.
- FIG. 6 is a flowchart of processing for one image in the present embodiment.
- step S101 pixel values are input.
- step S102 spatial noise removal processing is performed on the pixel value input in step S101.
- step S103 the probability that the pixel is stationary is calculated based on the pixel value from which spatial noise is removed in step S102, the pixel value input in step S101, and the pixel value of the past image.
- step S104 the time blend ratio is calculated based on the probability calculated in step S103.
- step S105 based on the time blend ratio calculated in step S104, the time blend process of the pixel value input in step S101 and the pixel value of the past image is performed.
- step S106 the output image after the time blending process in step S105 is performed is output.
- step S107 the output image is recorded in the buffer.
- step S108 it is determined whether all the pixels have been processed. If it is determined that the processing has been completed, the processing for the one image is terminated (step S109). If it is determined that the process has not ended, the process proceeds to step S101.
- the noise removal filter process can be performed on all the images constituting the moving image by sequentially performing the above processing on each of the images constituting the moving image.
- the image noise removal device of the present embodiment in addition to the second image and the first noise removal image, the spatial noise removal image obtained by spatially removing the noise of the second image is used.
- the image noise removal device can perform temporal noise removal processing with appropriate intensity on the image.
- the detection of motion in the second image according to the conventional technique is performed by using the second image and the first noise-removed image, and it is known that the detection result is inaccurate.
- the accuracy of the motion detection result can be improved by using the spatial noise-removed image of the second image in addition to the above two images.
- the smaller the movement in the second image the more time noise removal processing can be performed on the image with the weight of the first noise removal image being increased.
- the noise of the image after the temporal noise removal process can be reduced by increasing the weight of the first noise-removed image from which noise has already been removed.
- the image noise removal device increases the weight of the first noise-removed image in temporal noise removal when it is determined that there is less possibility of movement in the second image. Since the noise has already been removed from the first noise-removed image, the noise of the image after the temporal noise removal processing can be reduced.
- the image noise removing device can make the size determination using the ratio to the deviation of the distribution of the pixel value that changes due to the noise when determining the size of the difference between the pixel values.
- the image noise removing device has an angle formed between the first vector corresponding to the difference between the spatial noise removed image and the second image and the second vector corresponding to the difference between the first noise removed image and the second image. It is determined that there is a smaller possibility of movement in the second image as the value is smaller, and the weight of the first noise-removed image in temporal noise removal is increased. Since the noise has already been removed from the first noise-removed image, the noise of the image after the temporal noise removal processing can be reduced.
- the image noise removing device can use an image obtained by performing temporal noise removing processing on the first image by the image noise removing device as the first noise removed image.
- a time noise removal process is sequentially performed on successive images in consideration of a noise removal result of an image immediately before the image. Can do.
- FIG. 7 is a configuration diagram of the image noise removing apparatus according to the present embodiment.
- the image noise removing apparatus 2 includes an input image terminal 10, a spatial noise removing unit 20, a reliability calculating unit 30, a time blending rate calculating unit 40, a time blending unit 50, and a buffer. 60, an output image terminal 70, a spatial blend rate calculation unit 80, and a spatial blend unit 90.
- the input image terminal 10 receives an image (hereinafter referred to as “input image”) input to the image noise removal device 1. Specifically, the input image terminal 10 receives the pixel value (pixel data) of the target pixel and the pixel values (pixel data) of the surrounding pixels.
- the spatial noise removal unit 20 performs noise removal by smoothing using the pixel values received by the input image terminal 10 and the pixel values of surrounding pixels.
- the reliability calculation unit 30 calculates the probability that the target pixel received by the input image terminal 10 is stationary with respect to the pixel at the same position in the past input image. That is, the reliability calculation unit 30 receives the pixel value of the target pixel received by the input image terminal 10, the pixel value at the same position in the image obtained by removing noise from the past image recorded in the buffer, and the space. The pixel value at the same position in the image after the noise removal unit 20 performs noise removal is used as an input, and the pixel of interest is stationary with respect to the pixel at the same position in the noise removal signal for the past image using these values. Probability is calculated.
- the spatial blend ratio calculation unit 80 is based on the probability calculated by the reliability calculation unit 30 and the pixel value received by the input image terminal 10 used in the spatial blend unit 90 based on the probability that the target pixel is stationary.
- the blend ratio with the pixel value after the processing by the noise removing unit 20 is calculated.
- the spatial blend unit 90 blends the pixel value received by the input image terminal 10 and the pixel value at the same position in the image after processing by the spatial noise removal unit 20 according to the blend rate calculated by the spatial blend rate calculation unit 80. I do.
- the time blend rate calculation unit 40 Based on the probability that the target pixel calculated by the reliability calculation unit 30 is still, the time blend rate calculation unit 40 outputs the output image of the spatial blend unit 90 used by the time blend unit 50 and the past recorded in the buffer. A blend ratio (mixing ratio) with the pixel value at the same position in the noise removal signal for the image is calculated.
- the time blending unit 50 blends the image value output from the spatial blending unit 90 and the pixel value at the same position in the past image recorded in the buffer according to the blending ratio calculated by the time blending rate calculating unit 40. Thus, an output image is generated, and the generated output image is output.
- the spatial blend rate calculation unit 80 calculates the blend rate ⁇ snr based on the probability P that the target pixel described in the first embodiment is stationary.
- the blend rate ⁇ snr is calculated in consideration of the following, for example.
- the blend rate is adjusted so that the input image terminal 10 is blended at a relatively high rate for the purpose of suppressing the side effects such as the above-mentioned blur.
- ⁇ snr is normalized and output, but normalization is not necessarily required in relation to the calculation in the spatial blending unit 90 in the subsequent stage.
- ⁇ snr may be obtained by calculation using the probability P, or may be calculated using a table or correspondence prepared in advance. As described in the first embodiment, the correspondence shown in FIG. 5 may be adopted as the correspondence.
- the probability P normalized by 1 can be used as it is as the blend rate ⁇ snr normalized by 1 without considering the side effect of spatial noise removal.
- the space blend ratio calculation unit 80 can be omitted, and the configuration can be simplified.
- the temporal noise removal filter can be optimally applied to the target image by adaptively changing the filter strength of the temporal noise removal filter, and the space for the moving part can be optimized. Noise removal can be performed by smoothly switching to the noise removal filter.
- FIG. 8 is a flowchart showing the flow of processing in the present embodiment.
- FIG. 8 is a flowchart of processing for one image in the present embodiment.
- step S101 pixel values are input.
- step S102 spatial noise removal processing is performed on the pixel value input in step S101.
- step S103 the probability that the pixel is stationary is calculated based on the pixel value from which spatial noise is removed in step S102, the pixel value input in step S101, and the pixel value of the past image.
- step S201 the spatial blend ratio is calculated based on the probability calculated in step S103.
- step S202 based on the spatial blend ratio calculated in step S201, spatial blend processing is performed on the pixel value input in step S101 and the pixel value from which spatial noise is removed.
- step S104 the time blend ratio is calculated based on the probability calculated in step S103.
- step S105 based on the time blend ratio calculated in step S104, the time blend process of the pixel value input in step S101 and the pixel value processed in step S202 is performed.
- step S106 the output image after the time blending process in step 105 is performed is output.
- step S107 the output image is recorded in the buffer.
- step S108 it is determined whether all the pixels have been processed. If it is determined that the processing has been completed, the processing for the one image is terminated (step S109). If it is determined that the process has not ended, the process proceeds to step S101.
- noise removal filter processing that adaptively changes the strength of the temporal noise removal filter and gently switches the moving portion with the spatial noise removal filter.
- the motion in the second image is detected with high accuracy as described above, and the spatial noise removal is performed on the second image according to the detection result. Later, temporal noise removal can be performed. Since the accuracy of motion detection in the second image is high, the accuracy of spatial noise removal can be improved. Then, since the temporal noise removal is performed after the spatial noise removal, a more appropriate temporal noise removal process can be performed on the image.
- FIG. 9 is an example of a configuration diagram of an image noise removing device in each embodiment.
- an image noise removing apparatus 1A for removing noise included in the second image after the first image in time order includes a spatial noise removing unit 20, a reliability calculating unit 30, A time blending unit 50.
- the spatial noise removal unit 20 generates a spatial noise-removed image (SNR) by performing an operation for removing noise included in the second image (CUR) using the pixel values included in the second image (CUR). To do.
- SNR spatial noise-removed image
- the reliability calculation unit 30 uses the second image (CUR) based on the spatial noise-removed image (SNR), the second image (CUR), and the first noise-removed image (PRE) obtained by removing noise from the first image. ) To calculate the reliability, which is an index of the presence or absence of movement.
- the time blending unit 50 performs addition processing based on reliability on the second image (CUR) and the first noise-removed image (PRE).
- each component may be configured by dedicated hardware or may be realized by executing a software program suitable for each component.
- Each component may be realized by a program execution unit such as a CPU or a processor reading and executing a software program recorded on a recording medium such as a hard disk or a semiconductor memory.
- the software that realizes the image noise removing apparatus and the like of each of the above embodiments is the following program.
- this program is an image noise removal method for removing noise contained in the second image after removing noise from the first image, using the pixel value contained in the second image.
- a spatial noise removal step of generating a spatial noise-removed image by performing an operation for removing noise included in the second image, the spatial noise-removed image, the second image, and the noise of the first image A reliability calculation step for calculating a reliability of movement in the second image based on the first noise-removed image from which the image is removed, and the reliability for the second image and the first noise-removed image.
- an image noise removing method including a time blending step for removing noise contained in the second image is executed.
- the image noise removing apparatus according to one or a plurality of aspects has been described based on the embodiment, but the present invention is not limited to this embodiment. Unless it deviates from the gist of the present invention, various modifications conceived by those skilled in the art have been made in this embodiment, and forms constructed by combining components in different embodiments are also within the scope of one or more aspects. May be included.
- the technique of the image noise removal apparatus can adaptively change the strength of the temporal noise removal filter, which is useful for noise removal of digital video cameras and digital still cameras that capture moving images. .
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Picture Signal Circuits (AREA)
- Image Processing (AREA)
Abstract
Description
本発明者は、「背景技術」の欄において記載した、ノイズ除去技術に関し、以下の問題が生じることを見出した。
図1は、本実施の形態における画像ノイズ除去装置の構成図である。
図7は、本実施の形態における画像ノイズ除去装置の構成図である。
10 入力画像端子
20 空間ノイズ除去部
30 信頼度算出部
31a、31b、31c 差分演算部
32 確率算出部
40 時間ブレンド率算出部
50 時間ブレンド部
60 バッファ
70 出力画像端子
80 空間ブレンド率算出部
90 空間ブレンド部
Claims (13)
- 第一画像のノイズ除去の後に、第二画像に含まれるノイズを除去するための画像ノイズ除去装置であって、
前記第二画像に含まれる画素値を用いて前記第二画像に含まれるノイズを除去するための演算を行うことで空間ノイズ除去画像を生成する空間ノイズ除去部と、
前記空間ノイズ除去画像と、前記第二画像と、前記第一画像のノイズを除去した第一ノイズ除去画像とに基づいて、前記第二画像における動きについての信頼度を算出する信頼度算出部と、
前記第二画像と前記第一ノイズ除去画像とに対して、前記信頼度に基づいた重み付け加算処理を行うことで、前記第二画像に含まれるノイズを除去する時間ブレンド部とを備える
画像ノイズ除去装置。 - 前記信頼度算出部は、
前記第二画像に含まれる画素の動きが小さいほど、より大きな値を前記信頼度として算出し、
前記時間ブレンド部は、
前記信頼度が大きいほど、前記第一ノイズ除去画像の重みをより大きくした重み付け加算処理を、前記重み付け加算処理として行う
請求項1に記載の画像ノイズ除去装置。 - 前記信頼度算出部は、
前記第一ノイズ除去画像と前記第二画像とのそれぞれに含まれる同一位置の画素値の差分が小さいほど、より大きな値を前記信頼度として算出する
請求項2に記載の画像ノイズ除去装置。 - 前記信頼度算出部は、
前記第一ノイズ除去画像と前記空間ノイズ除去画像とのそれぞれに含まれる同一位置の画素値の差分が小さいほど、より大きな値を前記信頼度として算出する
請求項2又は3に記載の画像ノイズ除去装置。 - 前記信頼度算出部は、
予め定められた偏差であってノイズにより変化する画素値の分布の偏差に対する、前記差分の比に基づいて、前記信頼度を算出する
請求項3又は4に記載の画像ノイズ除去装置。 - 前記画素値は、複数の成分を有し、
前記信頼度算出部は、
前記空間ノイズ除去画像と前記第二画像とのそれぞれにおける同一位置の画素の画素値の差分を成分として有する第一ベクトルと、前記第一ノイズ除去画像と前記第二画像とのそれぞれにおける同一位置の画素の画素値の差分を成分として有する第二ベクトルとのなす角が小さいほど、より大きな値を前記信頼度として算出する
請求項2~5のいずれか1項に記載の画像ノイズ除去装置。 - 前記第一ノイズ除去画像は、前記第一画像に含まれるノイズを当該画像ノイズ除去装置により除去した画像である
請求項1~6のいずれか1項に記載の画像ノイズ除去装置。 - 前記第一画像は、動画像を構成する画像のうちの1つの画像であり、
前記第二画像は、前記動画像を構成する画像のうちの1つの画像であって、時間順で前記第一画像の直後の画像である
請求項1~7のいずれか1項に記載の画像ノイズ除去装置。 - 前記画像ノイズ除去装置は、さらに、
前記第二画像と前記空間ノイズ除去画像とに対して、前記信頼度に基づいた重み付け加算処理を行うことで、空間ブレンド画像を生成する空間ブレンド部とを備え、
前記時間ブレンド部は、さらに、
前記空間ブレンド画像と前記第一画像に含まれるノイズを除去した第一ノイズ除去画像とに対して、前記信頼度に基づいた重み付け加算処理を行うことで、前記第二画像に含まれるノイズを除去する
請求項1に記載の画像ノイズ除去装置。 - 前記信頼度算出部は、
前記第二画像に含まれる画素の動きが小さいほど、より大きな値を前記信頼度として算出し、
前記空間ブレンド部は、
前記信頼度が大きいほど、前記第二画像の重みをより大きくした重み付け加算処理を、前記重み付け加算処理として行い、
前記時間ブレンド部は、
前記信頼度が大きいほど、前記第一ノイズ除去画像の重みをより大きくした重み付け加算処理を、前記重み付け加算処理として行う
請求項1に記載の画像ノイズ除去装置。 - 第一画像のノイズ除去の後に、第二画像に含まれるノイズを除去するための画像ノイズ除去方法であって、
前記第二画像に含まれる画素値を用いて前記第二画像に含まれるノイズを除去するための演算を行うことで空間ノイズ除去画像を生成する空間ノイズ除去ステップと、
前記空間ノイズ除去画像と、前記第二画像と、前記第一画像のノイズを除去した第一ノイズ除去画像とに基づいて、前記第二画像における動きについての信頼度を算出する信頼度算出ステップと、
前記第二画像と前記第一ノイズ除去画像とに対して、前記信頼度に基づいた重み付け加算処理を行うことで、前記第二画像に含まれるノイズを除去する時間ブレンドステップとを含む
画像ノイズ除去方法。 - 請求項11に記載の画像ノイズ除去方法をコンピュータに実行させるためのプログラム。
- 第一画像のノイズ除去の後に、第二画像に含まれるノイズを除去するための集積回路であって、
前記第二画像に含まれる画素値を用いて前記第二画像に含まれるノイズを除去するための演算を行うことで空間ノイズ除去画像を生成する空間ノイズ除去部と、
前記空間ノイズ除去画像と、前記第二画像と、前記第一画像のノイズを除去した第一ノイズ除去画像とに基づいて、前記第二画像における動きについての信頼度を算出する信頼度算出部と、
前記第二画像と前記第一ノイズ除去画像とに対して、前記信頼度に基づいた重み付け加算処理を行うことで、前記第二画像に含まれるノイズを除去する時間ブレンド部とを備える
集積回路。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201380004125.5A CN104115482B (zh) | 2012-10-04 | 2013-10-01 | 图像噪声除去装置、以及图像噪声除去方法 |
US14/359,961 US9367900B2 (en) | 2012-10-04 | 2013-10-01 | Image noise removing apparatus and image noise removing method |
JP2014509025A JP6254938B2 (ja) | 2012-10-04 | 2013-10-01 | 画像ノイズ除去装置、および画像ノイズ除去方法 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2012-221988 | 2012-10-04 | ||
JP2012221988 | 2012-10-04 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2014054273A1 true WO2014054273A1 (ja) | 2014-04-10 |
Family
ID=50434617
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2013/005844 WO2014054273A1 (ja) | 2012-10-04 | 2013-10-01 | 画像ノイズ除去装置、および画像ノイズ除去方法 |
Country Status (4)
Country | Link |
---|---|
US (1) | US9367900B2 (ja) |
JP (1) | JP6254938B2 (ja) |
CN (1) | CN104115482B (ja) |
WO (1) | WO2014054273A1 (ja) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109791689A (zh) * | 2016-09-30 | 2019-05-21 | 哈德利公司 | 图像信号处理器偏置补偿噪声降低系统和方法 |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP6238521B2 (ja) * | 2012-12-19 | 2017-11-29 | キヤノン株式会社 | 3次元計測装置およびその制御方法 |
JP6135526B2 (ja) * | 2014-01-30 | 2017-05-31 | 株式会社リガク | 画像処理方法および画像処理装置 |
FR3018147B1 (fr) * | 2014-03-03 | 2016-03-04 | Sagem Defense Securite | Debruitage video optimise pour systeme multicapteur heterogene |
KR101913650B1 (ko) * | 2017-11-06 | 2018-10-31 | 크루셜텍 (주) | 디스플레이 영역에서의 생체 이미지 판독 장치 |
US10469749B1 (en) * | 2018-05-01 | 2019-11-05 | Ambarella, Inc. | Temporal filter with criteria setting maximum amount of temporal blend |
CN110796624B (zh) | 2019-10-31 | 2022-07-05 | 北京金山云网络技术有限公司 | 一种图像生成方法、装置及电子设备 |
EP3832591B1 (en) * | 2019-12-03 | 2022-06-08 | Axis AB | Encoding of a video sequence |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007060636A (ja) * | 2005-07-29 | 2007-03-08 | Victor Co Of Japan Ltd | ノイズ検出装置及び方法、並びにノイズ低減装置及び方法 |
JP2010147840A (ja) * | 2008-12-19 | 2010-07-01 | Sanyo Electric Co Ltd | 画像処理装置及び撮像装置 |
JP2010147985A (ja) * | 2008-12-22 | 2010-07-01 | Sony Corp | 画像処理装置、および画像処理方法、並びにプログラム |
WO2010073488A1 (ja) * | 2008-12-22 | 2010-07-01 | パナソニック株式会社 | 画像ノイズ除去装置及び方法 |
Family Cites Families (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2782766B2 (ja) | 1989-03-22 | 1998-08-06 | キヤノン株式会社 | 動画静止画変換方法 |
GB9607668D0 (en) * | 1996-04-12 | 1996-06-12 | Snell & Wilcox Ltd | Video noise reducer |
US7295616B2 (en) * | 2003-11-17 | 2007-11-13 | Eastman Kodak Company | Method and system for video filtering with joint motion and noise estimation |
US7199838B2 (en) * | 2004-06-17 | 2007-04-03 | Samsung Electronics Co., Ltd. | Motion adaptive noise reduction apparatus and method for video signals |
US7769089B1 (en) * | 2004-12-02 | 2010-08-03 | Kolorific, Inc. | Method and system for reducing noise level in a video signal |
ATE530015T1 (de) * | 2005-01-18 | 2011-11-15 | Lg Electronics Inc | Anordnung zur entfernung von rauschen aus einem videosignal |
US7535517B2 (en) * | 2005-04-14 | 2009-05-19 | Samsung Electronics Co., Ltd. | Method of motion compensated temporal noise reduction |
TWI324013B (en) * | 2006-02-22 | 2010-04-21 | Huper Lab Co Ltd | Video noise reduction method using adaptive spatial and motion-compensation temporal filters |
JP4859632B2 (ja) * | 2006-11-15 | 2012-01-25 | 富士通セミコンダクター株式会社 | 画像処理装置及び画像処理方法 |
JP5052319B2 (ja) * | 2007-12-17 | 2012-10-17 | オリンパス株式会社 | 動画ノイズ低減処理装置、動画ノイズ低減処理プログラム、動画ノイズ低減処理方法 |
US20090161756A1 (en) * | 2007-12-19 | 2009-06-25 | Micron Technology, Inc. | Method and apparatus for motion adaptive pre-filtering |
US8184705B2 (en) * | 2008-06-25 | 2012-05-22 | Aptina Imaging Corporation | Method and apparatus for motion compensated filtering of video signals |
JPWO2010007777A1 (ja) * | 2008-07-16 | 2012-01-05 | パナソニック株式会社 | 画像処理装置、画像処理方法、プログラム、記録媒体および集積回路 |
US20100045870A1 (en) * | 2008-08-25 | 2010-02-25 | Mediatek Inc. | Adaptive noise reduction system |
CN101448077B (zh) * | 2008-12-26 | 2010-06-23 | 四川虹微技术有限公司 | 一种自适应视频图像3d降噪方法 |
TWI390467B (zh) * | 2009-01-23 | 2013-03-21 | Silicon Integrated Sys Corp | 動態雜訊濾波裝置及西格瑪濾波方法 |
CN101964863B (zh) * | 2010-05-07 | 2012-10-24 | 镇江唐桥微电子有限公司 | 一种自适应的时空域视频图像降噪方法 |
US8427583B2 (en) * | 2010-09-30 | 2013-04-23 | Sharp Laboratories Of America, Inc. | Automatic parameter control for spatial-temporal filter |
-
2013
- 2013-10-01 US US14/359,961 patent/US9367900B2/en not_active Expired - Fee Related
- 2013-10-01 JP JP2014509025A patent/JP6254938B2/ja not_active Expired - Fee Related
- 2013-10-01 WO PCT/JP2013/005844 patent/WO2014054273A1/ja active Application Filing
- 2013-10-01 CN CN201380004125.5A patent/CN104115482B/zh not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007060636A (ja) * | 2005-07-29 | 2007-03-08 | Victor Co Of Japan Ltd | ノイズ検出装置及び方法、並びにノイズ低減装置及び方法 |
JP2010147840A (ja) * | 2008-12-19 | 2010-07-01 | Sanyo Electric Co Ltd | 画像処理装置及び撮像装置 |
JP2010147985A (ja) * | 2008-12-22 | 2010-07-01 | Sony Corp | 画像処理装置、および画像処理方法、並びにプログラム |
WO2010073488A1 (ja) * | 2008-12-22 | 2010-07-01 | パナソニック株式会社 | 画像ノイズ除去装置及び方法 |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109791689A (zh) * | 2016-09-30 | 2019-05-21 | 哈德利公司 | 图像信号处理器偏置补偿噪声降低系统和方法 |
KR20190054126A (ko) * | 2016-09-30 | 2019-05-21 | 허들리 인코포레이티드 | Isp 바이어스-보상 노이즈 감소 시스템들 및 방법들 |
JP2019530360A (ja) * | 2016-09-30 | 2019-10-17 | ハドリー インコーポレイテッド | Ispバイアス補償ノイズ低減システムおよび方法 |
KR102411033B1 (ko) * | 2016-09-30 | 2022-06-17 | 허들리 인코포레이티드 | Isp 바이어스-보상 노이즈 감소 시스템들 및 방법들 |
Also Published As
Publication number | Publication date |
---|---|
JPWO2014054273A1 (ja) | 2016-08-25 |
JP6254938B2 (ja) | 2017-12-27 |
CN104115482B (zh) | 2017-11-21 |
CN104115482A (zh) | 2014-10-22 |
US9367900B2 (en) | 2016-06-14 |
US20140341480A1 (en) | 2014-11-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6254938B2 (ja) | 画像ノイズ除去装置、および画像ノイズ除去方法 | |
JP4727720B2 (ja) | 画像処理方法および画像処理装置 | |
US9413951B2 (en) | Dynamic motion estimation and compensation for temporal filtering | |
US9514525B2 (en) | Temporal filtering for image data using spatial filtering and noise history | |
US9135683B2 (en) | System and method for temporal video image enhancement | |
US9727984B2 (en) | Electronic device and method for processing an image | |
JP2007274299A5 (ja) | ||
JP2013508811A5 (ja) | ||
US20100067818A1 (en) | System and method for high quality image and video upscaling | |
US20100111438A1 (en) | Anisotropic diffusion method and apparatus based on direction of edge | |
JP2012208553A (ja) | 画像処理装置、および画像処理方法、並びにプログラム | |
JPWO2017169039A1 (ja) | 画像処理装置、撮像装置、および画像処理方法、並びにプログラム | |
JP2009212969A (ja) | 画像処理装置、画像処理方法、及び画像処理プログラム | |
JP4935930B2 (ja) | 画像補正装置および画像補正方法 | |
US8750635B2 (en) | Image processing apparatus, image processing method, program, and recording medium | |
JP2012073703A (ja) | 画像ボケ量計算装置およびそのプログラム | |
JP6576179B2 (ja) | 画像処理装置、画像処理方法、及びプログラム | |
JP2016201037A (ja) | 画像処理装置、画像処理方法及びプログラム | |
Zhao et al. | An improved image deconvolution approach using local constraint | |
JP3959547B2 (ja) | 画像処理装置、画像処理方法、及び情報端末装置 | |
JP5701035B2 (ja) | 画像処理装置および方法並びにプログラム | |
US20160328858A1 (en) | Image processing system, image processing method, and image processing program | |
JP4930638B2 (ja) | 画像補正装置および画像補正方法 | |
JP5868197B2 (ja) | ノイズ低減装置およびその制御方法 | |
JP2014158083A5 (ja) |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
ENP | Entry into the national phase |
Ref document number: 2014509025 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 14359961 Country of ref document: US |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 13844399 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 13844399 Country of ref document: EP Kind code of ref document: A1 |