JP2009219066A - Image processing apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processing apparatus for reducing noise of an entire image including an edge peripheral portion while preventing an edge portion in the image from being blurred. <P>SOLUTION: The image processing apparatus 1000 includes: a wavelet transformation section 110 which performs wavelet transformation upon an input image signal to separate the input image signal into a high-frequency signal component and a low-frequency signal component; and a time integration type filter section 120 which performs time integration processing upon the high-frequency signal component to cancel noise in a time-base direction from the high-frequency signal component. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an image processing apparatus that reduces image noise using multi-resolution conversion such as wavelet conversion.

  Conventionally, as a technique for reducing noise in a video signal, a noise reduction technique using a wavelet transform (Wavelet Transform) which is one of multi-resolution conversions is known. The noise reduction technology using wavelet transform reduces image noise as follows.

  First, using wavelet transform, the image signal is separated into a low frequency band including a low-frequency signal component and a high-frequency band subband including a high-frequency signal component. Similar processing is repeated. As a result, a frequency band image signal (hereinafter referred to as a subband image signal) obtained by separating the image signal into subbands for each frequency band is generated.

  Next, a coring process for removing a signal component having a minute amplitude value is performed on the subband image signal in the high frequency band including the high frequency signal component. Then, wavelet inverse transformation is performed by synthesizing a plurality of separated subband image signals including the subband image signal of the high-frequency signal component after coring processing. Thereby, an image signal in which only a noise component with a small amplitude included in the high frequency signal component is suppressed is reconstructed.

  According to such noise reduction technology, noise having a specific frequency component can be reduced by selecting a high frequency band for performing coring processing, or noise reduction can be achieved by changing the conditions of coring processing. Can be changed, and fine noise reduction processing can be performed.

  Daubechies bases and Haar bases are known as specific methods of wavelet transform. The base of Daubechies has high band separation accuracy, but the filter coefficients are complicated. On the other hand, the Haar base can be easily configured with only two filter coefficients “1” and “−1”, although the accuracy of band separation is low. For general civilian equipment, the Haar base, which requires a small hardware scale, is used.

  FIG. 13 shows a basic configuration of a conventional image processing apparatus that reduces noise based on the Haar base. Here, an example is shown in which the number of times of separation into subbands is two. Hereinafter, the number of times of separation into subbands is called a level. That is, the input image signal is set to level 0, and the result of performing one subband separation on this input image signal is called level 1, and another subband separation is performed on this result. The result is called level 2.

  As illustrated in FIG. 13, the image processing device 5000 includes a wavelet transform unit 710, a coring processing unit 740, and a wavelet inverse transform unit 730. When the image signal is input to the wavelet transform unit 710, the wavelet transform unit 710 performs wavelet transform. Thereby, the input image signal is separated into subbands for each frequency band, and a plurality of subband image signals are generated. At this time, as the plurality of subband image signals, a low frequency band including a low frequency signal component of the image signal and one or more high frequency bands including a high frequency signal component are generated.

  The coring processing unit 740 removes noise contained in the high frequency signal component by setting the minute amplitude symbol to 0 for the high frequency signal component obtained by the wavelet transform unit 710. The wavelet inverse transform unit 730 performs wavelet inverse transform on the high-frequency signal component after the coring process and the low-frequency signal component converted by the wavelet transform unit 120. Thereby, the image signal is reconstructed and an image from which noise is removed is output.

  FIG. 14 is a diagram for explaining the operation of the conventional image processing apparatus 5000. As shown in FIG. 14A, in this example, the right side of the input image is bright and the left side is dark, and there is noise throughout. FIG. 14B shows the signal amplitude in the X direction of the input image shown in FIG. FIG. 14C shows one of the high frequency signal components obtained by converting the input image by the wavelet transform unit 710, and the high frequency signal component has noise as shown in the figure. Therefore, by performing coring processing on the high frequency signal component by the coring processing unit 740, noise in the high frequency signal component is removed.

  However, when coring processing is performed to uniformly reduce the amplitude of noise with respect to the amplitude of the high-frequency signal component, the edge portion M (the boundary between the bright region and the dark region of the input image) in FIG. As a result, the high-frequency signal component is also reduced, and the edge portion of the output image may be blurred.

Therefore, Patent Document 1 determines whether or not each position of the high-frequency signal component to be coring corresponds to the edge portion M from the signal value at the corresponding position of the high-frequency signal component at a higher level than the coring target. Thus, the coring threshold value of the edge portion M is set smaller than the threshold values other than the edge portion. FIG. 15A shows the result, and noise other than the edge peripheral portion N is removed while preventing a decrease in the high frequency signal component of the edge peripheral portion N.
JP 2003-134352 A (particularly paragraph 0068)

  However, in the conventional image signal processing apparatus, as shown in FIG. 15 (a), since the coring threshold is lowered at the edge peripheral portion N including the edge portion M, the edge portion peripheral portion N There was a problem that noise remained without being removed. FIG. 15B is an output image corresponding to FIG. 15A and shows noise remaining in the edge peripheral portion N. FIG.

  The present invention has been made to solve the above-described conventional problems, and provides an image processing apparatus capable of reducing noise in the entire image including the periphery of the edge portion while preventing the edge portion in the image from blurring. The purpose is to provide.

  An image processing apparatus according to the present invention includes a subband converting unit that performs subband conversion on an input image signal to separate the input image signal into a high frequency signal component and a low frequency signal component; and the high frequency signal A time integration type filter unit that removes noise in the time axis direction from the high-frequency signal component by performing time integration processing on the component is provided.

  With this configuration, time integration processing is performed on the high frequency signal component. Since the noise of the image signal has a characteristic that changes in the time axis direction, the noise can be removed from the high-frequency signal component by performing the time integration process. In this time integration processing, noise around the edge is also removed in the same manner as other portions. Further, even if time integration processing is performed, the signal value at the edge portion does not decrease. Therefore, it is possible to reduce noise of the entire image including the periphery of the edge portion while preventing the edge portion in the image from blurring.

  Further, the image processing apparatus according to the present invention provides an edge for creating an edge signal obtained by performing subband inverse transform only on the high-frequency signal component among the high-frequency signal component and the low-frequency signal component. A signal creation unit, wherein the time integration type filter unit performs time integration processing using the edge signal as the high frequency signal component, and further, the edge signal processed by the time integration type filter unit, and the low frequency range An edge signal adding unit that adds signal components is provided.

  With this configuration, an edge signal obtained by performing subband inverse transformation only on the high frequency signal component is created, and time integration processing is performed using the edge signal as the high frequency signal component. Therefore, the high frequency signal component can be processed by one time integration type filter unit, and the configuration can be simplified.

  In the image processing apparatus according to the present invention, the time integration type filter unit performs time integration processing on a signal component of a minute amplitude value in a filter input signal, the signal after time integration processing, and the filter input It has a configuration in which a signal subjected to coring processing for removing a signal component having a minute amplitude value in the signal is added.

  With this configuration, the time integration filter unit performs time integration processing on a signal component having a minute amplitude value in the filter input signal. As a result, since the time integration process is not applied to signal components other than the signal component having a minute amplitude value, it is possible to prevent blurring from occurring at the edge portion of the moving object.

  The image processing apparatus according to the present invention includes a coring processing unit that removes a signal component having a minute amplitude value in the high frequency signal component, a high frequency signal component obtained by the coring processing unit, and the low frequency signal. The subband inverse transform unit that performs subband inverse transform on the signal component, the minute signal extraction unit that extracts the signal component of the minute amplitude value in the high frequency signal component, and the minute signal extraction unit A minute edge signal creating unit that creates an edge signal having a minute amplitude value obtained by performing subband inverse transformation on a high frequency signal component having a minute amplitude value, and the time integrating filter unit includes the minute amplitude signal Time integration processing using the edge signal of the value as the high-frequency signal component, and further, the edge signal of the minute amplitude value processed by the time integration filter unit and the signal obtained by the subband inverse conversion unit It has a configuration in which the edge signal adder for calculation is provided.

  With this configuration, a minute amplitude value signal component is extracted from the high frequency signal component, and an edge signal with a small amplitude value obtained by performing subband inverse transformation on the high frequency signal component of the small amplitude value is created. Then, time integration processing is performed using an edge signal having a small amplitude value as a high-frequency signal component. As a result, since the time integration process is not applied to signal components other than the signal component having a minute amplitude value, it is possible to prevent blurring from occurring at the edge portion of the moving object, and to convert the edge signal into time. Since the integration process is performed, the high-frequency signal component can be processed by one time integration filter, and the configuration can be simplified.

  In the image processing apparatus according to the present invention, the time integration filter unit includes a cyclic filter that adds a filter input signal and a filtered signal of the previous frame at a predetermined ratio. ing. Thereby, noise removal by time integration processing can be suitably performed.

  In the image processing apparatus according to the present invention, the time integration filter unit has a configuration for performing a coring process for removing a signal component having a minute amplitude value in a signal processed by the cyclic filter. Thereby, the influence of the past noise which a recursive filter outputs can be reduced.

  In the image processing apparatus according to the present invention, the time integration type filter unit is configured to be able to change a time constant of the time integration process. Thereby, the effect of time integration processing can be adjusted and noise removal can be suitably performed.

  The image processing apparatus according to the present invention further includes a motion detection unit that detects a motion of the input image signal, and the time integration filter unit includes the time constant at a position where the motion is detected by the motion detection unit. It has the structure which changes. Thereby, it is possible to prevent blurring from occurring at the edge portion of the moving object.

  In the image processing apparatus according to the present invention, the motion detection unit has a configuration for detecting the motion by detecting a difference between the input image signals. Thereby, the motion of the input image signal can be suitably detected.

  The present invention removes noise in the time axis direction from the high-frequency signal component by performing time integration processing on the high-frequency signal component, thereby preventing the edge portion in the image from blurring. It is possible to provide an image processing apparatus having an effect that noise of the entire image including the periphery can be reduced.

(First embodiment)
Hereinafter, an image processing apparatus according to a first embodiment of the present invention will be described with reference to the drawings.

  FIG. 1 shows the configuration of an image processing apparatus 1000 according to the first embodiment of the present invention. In FIG. 1, the image processing apparatus 1000 includes a wavelet transform unit 110, a time integration filter unit 120, and a wavelet inverse transform unit 130. In the example of the present embodiment, the number of times of separation into subbands is two.

  The wavelet transform unit 110 performs wavelet transform on an input image signal, thereby including a low frequency band including a low frequency signal component (low frequency signal component) and a high frequency signal component (high frequency signal component). It converts into the subband image signal for every frequency band isolate | separated into the above high frequency band. The wavelet transform unit 110 is an example of a subband transform unit of the present invention, and the wavelet transform is an example of a subband transform.

  The wavelet transform unit 110 includes a level 1 transform unit 111 and a level 2 transform unit 112. The level 1 converter 111 performs the first (level 1) wavelet transform on the input image signal. The level 2 conversion unit 112 performs a second (level 2) wavelet transform on the image signal after the first wavelet transform.

  The time integration type filter unit 120 removes noise in the time axis direction from the high frequency signal component by performing time integration processing on the high frequency signal component obtained by the wavelet transform unit 110. The time integration type filter unit 120 includes a first high frequency signal time integration unit 121 and a second high frequency signal time integration unit 122. The first high frequency signal time integration unit 121 performs time integration processing on the high frequency signal component obtained by the level 1 conversion unit 111. The second high frequency signal time integration unit 122 performs time integration processing on the high frequency signal component obtained by the level 2 conversion unit 112.

  The wavelet inverse transform unit 130 performs wavelet inverse transform on the high frequency signal component from which noise has been removed by the time integration filter unit 120 and the low frequency signal component obtained by the wavelet transform unit 110. By this wavelet inverse transform, the high-frequency signal component from which noise has been removed and the low-frequency signal component are combined to reconstruct an image signal, which becomes an output image signal. The wavelet inverse transform unit 130 is an example of a subband inverse transform unit that synthesizes subband image signals, and the wavelet inverse transform is an example of a subband inverse transform.

  The wavelet inverse transform unit 130 includes a level 1 inverse transform unit 131 and a level 2 inverse transform unit 132. The level 2 inverse transform unit 132 performs the high frequency signal component from which the noise in the time axis direction has been removed by the second high frequency signal time integration unit 122 and the low frequency signal component obtained by the level 2 conversion unit 112. Perform inverse wavelet transform. The level 1 inverse transform unit 131 uses the wavelet for the signal component obtained by the level 2 inverse transform unit 132 and the high frequency signal component from which noise in the time axis direction is removed by the first high frequency signal time integration unit 121. Perform inverse transformation.

Next, the operation of the image processing apparatus 1000 according to the first embodiment of the present invention will be described with reference to the drawings.
With the configuration shown in FIG. 1, the image processing apparatus 1000 performs a process of reducing noise included in an image signal using wavelet transform, which is one of multi-resolution conversions. In the present embodiment, subband conversion is performed in the frequency band of the horizontal frequency and the vertical frequency based on the Haar base. The number of separation into subbands is two.

  First, the procedure of wavelet transformation by the wavelet transformation unit 110 will be described. FIG. 2A is a diagram showing the acquisition position of each sample of the input image signal for converting the input image signal into subbands up to level 2. FIG. Signals S00 to S33 indicate input image signals delayed in pixel units, and the pixel signals S00 to S33 are arranged in a matrix. Each input image signal is delayed by one pixel in the horizontal direction from signal S00 to signal S03, and delayed by one pixel in the vertical direction from signal S00 to signal S33.

FIG. 2B is a diagram showing a subband separation state after the input image signal is converted into subbands up to level 2. FIG. In FIG. 2B, numbers such as “1” and “2” indicate wavelet transform levels, “L” indicates a low frequency band, and “H” indicates a high frequency band. Also, the frequency F _H of the horizontal direction is high-frequency low-frequency, at the right in the left, the frequency F _V of the vertical direction is high-frequency low-band at the upper, at the lower side.

By performing wavelet transform on the input image signal, the input image signal is converted into a horizontal low frequency vertical low frequency “ ₁ LL”, a horizontal high frequency vertical low frequency “ ₁ HL”, and a horizontal low frequency vertical high frequency “ ₁ LH”. And a horizontal high band and a vertical high band “ ₁ HH”. Here, “ ₁ LL” is a low-frequency signal component, and “ ₁ HL”, “ ₁ LH”, and “ ₁ HH” are high-frequency signal components. In the level 2 wavelet transform, the wavelet transform is further performed on the low-frequency signal component “ ₁ LL” after the level 1 wavelet transform. As a result, as shown in FIG. 2 (b), “ ₁ LL” is the horizontal low frequency vertical low frequency “ ₂ LL”, the horizontal high frequency vertical low frequency “ ₂ HL”, and the horizontal low frequency vertical high frequency “ ₂ LH”. ”And the horizontal high band vertical high band“ ₂ HH ”. Similarly to the above, “ ₂ LL” is a low-frequency signal component, and “ ₂ HL”, “ ₂ LH”, and “ ₂ HH” are high-frequency signal components.

  FIG. 3 is a diagram illustrating a procedure of wavelet transform performed by the wavelet transform unit 110. Here, referring to FIG. 3, a converted signal for each frequency band separated by the sub-band shown in FIG. 2B is generated from each sample of the input image signal shown in FIG. The procedure to do is explained.

As shown in the equation (1), the wavelet transform in the Haar base is performed by using a level 0 low frequency signal (each sample value in FIG. 3A) as an input image signal from a level n low frequency band to a level n + 1. It is expressed in the form of a recurrence formula for converting into each frequency component. Regarding the sign on the right side of “LL” in Expression (1), “V” represents the vertical position and “H” represents the horizontal position, and the conversion signal of the vertical position V and the horizontal position H is represented by Expression (1). (The conversion symbol indicates a subband image signal subjected to subband conversion, and so on).

FIG. 3A is a diagram showing a low-frequency signal of level 0 (each sample of the input image signal) in Expression (1), and is the same as FIG. 2A. FIG. 3B is a diagram showing the result after the level 1 wavelet transform. The level 1 conversion unit 111 of the wavelet conversion unit 110 performs the processing of Expression (1) on the low-frequency signal of level 0. As a result, the low-frequency signal component ₀ LL of the input image signal is converted into the converted signal ₁ LL of level 1 Converted to ₁ HL, ₁ LH and ₁ HH. Level 1 low-frequency signal components ₁ LL _0,0 , ₁ LL _0,1 , ₁ LL _1,0 and ₁ LL _1,1 are extracted, and these extracted components are level 2 wavelets described below. Used for conversion.

FIG. 3C is a diagram showing the result after the level 2 wavelet transform. The level 2 conversion unit 112 performs processing of the expression (1) on the low frequency signal component ₁ LL obtained by the level 1 wavelet transform, and as a result, the level 1 low frequency signal component ₁ LL is the level 2 converted signal _2. Converted to LL, ₂ HL, ₂ LH and ₂ HH. The position of each signal after conversion is down-converted to ½, so that the position shown in FIG.

Among the level 2 converted signals obtained by the wavelet transform, the low-frequency signal component ₂ LL _0,0 is input to the level 2 inverse transform unit 132 of the wavelet inverse transform unit 130 (hereinafter, the vertical position V = 0, The processing of the present invention will be described focusing on the signal at the horizontal position H = 0, but it goes without saying that signals at other positions are also processed in the same manner.) On the other hand, the high frequency signal components ₂ HL _0,0 , ₂ LH _0,0 and ₂ HH _0,0 are respectively input to the second high frequency signal time integration unit 122 of the time integration type filter unit 120. Further, among the level 1 converted signals obtained by the wavelet transform, the high frequency signal components ₁ HL _0,0 , ₁ LH _0,0 and ₁ HH _0,0 are respectively the first high frequency of the time integrating filter unit 120. The signal is input to the signal time integration unit 121.

Next, the time integration filter unit 120 will be described in detail with reference to FIGS. As described with reference to FIG. 1, the time integration filter unit 120 includes the high-frequency signal components ₁ HL _0,0 , ₁ LH _0,0 , ₁ HH _0,0 , ₂ HL _0,0 , ₂ LH _0, Noise is removed from each high-frequency signal component by individually performing time integration processing on each of ₀ , ₂ HH _0,0 . The first high frequency signal time integration unit 121 of the time integration type filter unit 120 is a level 1 high frequency signal component ₁ HL _0,0 , ₁ LH _0,0 , ₁ HH _0,0 obtained by the level 1 conversion unit 111. Is subjected to time integration processing, and noise in the time axis direction is removed from the high-frequency signal component of level 1. The second high frequency signal time integration unit 122 performs time integration processing on the level 2 high frequency signal components ₂ HL _0,0 , ₂ LH _0,0 , ₂ HH _0,0 obtained by the level 2 conversion unit 112. The noise in the time axis direction is removed from the high frequency signal component of level 2.

  FIG. 4 is a diagram showing a configuration of a cyclic (Infinite Impulse Response: hereinafter referred to as IIR) filter 1200 as a specific example of the first high-frequency signal time integration unit 121 and the second high-frequency signal time integration unit 122. The IIR filter 1200 has a configuration in which an input signal (filter input signal) to the IIR filter 1200 and a signal subjected to filter processing in the previous frame are added at a predetermined ratio.

  As shown in FIG. 4, the IIR filter 1200 includes a frame delay memory 1201, multipliers 1202 a and 1202 b, and an adder 1203. The frame delay memory 1201 stores a high-frequency signal component after noise removal that has been filtered one frame before. When the time constant that is a parameter of the IIR filter 1200 is k (0 <k <1), the high-frequency signal component that is the filter input signal is multiplied by (1−k) by the multiplier 1202a and stored in the frame delay memory 1201. The high-frequency signal component after noise removal is multiplied by k by the multiplier 1202b.

  The high frequency signal components multiplied by the multipliers 1202a and 1202b are added by the adder 1203. That is, the adder 1203 uses the high-frequency signal component at a ratio of (high-frequency signal component which is a filter input signal) :( high-frequency signal component filtered one frame before) = (1−k): k. The components are added.

  In this manner, the IIR filter 1200 adds the high-frequency signal component that is the filter input signal and the high-frequency signal component that has been filtered one frame before at a predetermined ratio ((1-k): k). Thus, time integration processing is performed on the high-frequency signal component that is the filter input signal, and noise in the time axis direction is removed. It is also possible to store a signal that has been subjected to filter processing in the frame delay memory two or more frames before, and perform filter processing using this signal.

5 and 6 show noise removal processing by the IIR filter 1200 of FIG. FIG. 5A is a diagram illustrating an example of an input image signal including noise. The input image is bright on the right side and dark on the left side as in the image shown in FIG. FIG. 5B shows one of the high frequency signal components obtained by converting the input image signal by the wavelet transform unit 110. Here, the high-frequency signal component ₁ HH _0,0 is illustrated. In FIG. 5B, M indicates the position of the boundary portion (edge portion) between the bright region and the dark region of the input image, and P indicates the edge peripheral portion.

FIG. _6A shows changes in the signal strength in the time axis direction at the edge part M and the edge peripheral part P of the high-frequency signal component ₁ HH _0,0 shown in FIG. As shown in FIG. 6A, the high-frequency signal component ₁ HH _0,0 in the input image signal varies in the time axis direction due to noise. Therefore, when integration in the time axis direction is performed using the time integration type filter unit 120, as shown in FIG. 6B, time due to noise is present in both the edge part M and the edge peripheral part P of the high frequency signal component. There is no fluctuation.

FIG. 6C shows the signal strength of the high-frequency signal component after noise removal, and by removing noise in the time axis direction, the edge peripheral portion P is reduced without reducing the signal strength of the edge portion M. Noise can be removed from the entire image including Here, the noise removal procedure for the high-frequency signal component ₁ HH _{0,0 at} level 1 has been described, but ₁ HL _0,0 , ₁ LH _0,0 at level ₁ , ₂ HL _{0 at} level 2 _, The same processing is performed on ₀ , ₁ LH _0,0 and ₁ HH _0,0 to remove noise in the time axis direction.

In the following description, “′” is added to the high frequency signal component after noise removal. That is, the high-frequency signal component of level 1 after noise removal is expressed as ₁ HL ′ _0,0 , ₁ LH ′ _0,0 and ₁ HH ′ _0,0, and the high-frequency signal component of level 2 after noise removal is ₂ HL ′ _0,0 , ₂ LH ′ _0,0 and ₂ HH ′ _0,0 . Level 1 high-frequency signal components ₁ HL ′ _0,0 , ₁ LH ′ _0,0 and ₁ HH ′ _0,0 after noise removal are respectively input to the level 1 inverse transform unit 131 and level 2 after noise removal The high frequency signal components ₂ HL ′ _0,0 , ₂ LH ′ _0,0 and ₂ HH ′ _0,0 are input to the level 2 inverse transform unit 132, respectively.

  Next, a processing procedure performed by the wavelet inverse transform unit 130 will be described with reference to FIG.

The inverse wavelet transform in the Haar basis is expressed in the form of a recurrence formula for transforming each frequency band at level n + 1 to a low frequency band at level n, as shown in equation (2). Regarding the code on the right side of “LL”, “V” represents a vertical position and “H” represents a horizontal position.

FIG. 7A shows a level 2 conversion signal input to the wavelet inverse conversion unit 130. As described above, the low-frequency signal component ₂ LL _0,0 is input from the wavelet transform unit 110 to the level 2 inverse transform unit 132 of the wavelet inverse transform unit 130 and the level after noise removal from the time integration filter unit 120. Two high-frequency signal components ₂ HL ′ _0,0 , ₂ LH ′ _0,0 and ₂ HH ′ _0,0 are input.

FIG. 7B is a diagram showing the result after the level 2 wavelet inverse transform. The level 2 inverse transform unit 132 performs level 2 low-frequency signal component ₂ LL _0,0 and high-frequency signal component ₂ HL ′ _0,0 , ₂ LH ′ _0,0 , ₂ HH ′ _0,0 after noise removal. Then, the wavelet inverse transformation of the expression (2) is performed to inversely transform the level 1 low-frequency signal component ₁ LL ′ _0,0 .

FIG. 7C is a diagram showing the result after the inverse wavelet transform of level 1. The level 1 inverse transform unit 131 includes a level 1 low frequency signal component ₁ LL ′ _0,0 obtained by the level 2 inverse transform unit 132 and a level 1 high frequency signal component ₁ HL ′ _0,0 after noise removal. ₁ LH ′ _0,0 and ₁ HH ′ _0,0 are inversely transformed into level 0 low-frequency signal component ₀ LL ′ _0,0 by performing the wavelet inverse transformation of equation (2). Is output as an image signal after noise removal. Then, by performing inverse transformation of all components according to Expression (2), it is possible to obtain image signals S00 ′, S01 ′,..., S33 ′ after noise removal.

  FIG. 8A shows the signal intensity of the output image signal after noise removal. As shown in FIG. 8A, it can be seen that noise is removed from the output image signal compared to the input image signal shown in FIG. FIG. 8B shows an output image after noise removal. As shown in FIG. 8B, the edge portion M in the output image is not blurred, and the noise is also removed from the edge peripheral portion P.

  According to the image processing apparatus of the first embodiment of the present invention, time integration processing is performed on the high frequency signal component. Since the noise of the image signal has a characteristic that changes in the time axis direction, the noise can be removed from the high-frequency signal component by performing the time integration process. In this time integration processing, noise around the edge is also removed in the same manner as other portions. Further, even if time integration processing is performed, the signal value at the edge portion does not decrease. Therefore, it is possible to reduce noise of the entire image including the periphery of the edge portion while preventing the edge portion in the image from blurring.

  Further, according to the image processing apparatus of the first embodiment of the present invention, the time integration type filter unit adds the filter input signal and the filtered signal of the previous frame at a predetermined ratio. It has a configuration including a filter. Thereby, noise removal by time integration processing can be suitably performed.

  Next, a modification of the image processing apparatus according to the first embodiment of the present invention will be described. In the above-described embodiment, the first high frequency signal time integration unit 121 and the second high frequency signal time integration unit 122 are configured only by the IIR filter 1200 shown in FIG. In this case, since the time integration process is performed on all signals regardless of the amplitude of the high frequency signal component, blurring may occur at the edge portion of the moving object. In the modification described below, blurring at the edge of a moving object is reduced by performing noise removal only on a signal component with a small amplitude in the time axis direction.

  FIG. 9 shows a configuration of the first high frequency signal time integration unit 121 and the second high frequency signal time integration unit 122 in the present modification. As shown in FIG. 9, the present modification includes an IIR filter 1200, a coring filter (coring processing unit) 1204, a minute signal extraction unit 1205, and an addition unit 1203b. The configuration of the IIR filter 1200 is the same as the configuration shown in FIG.

  The coring filter 1204 has a configuration in which a component equal to or lower than the threshold value in the input signal is set to 0, thereby performing a process of removing a signal component having a minute amplitude value from a high-frequency signal component that is a filter input signal. FIG. 10A is a diagram showing an example of input / output characteristics of the coring processing unit 1204, and the output of the minute amplitude value is set to 0. As shown in FIG. 10A, when the value of the filter input signal is −th1 to th1, the signal component is replaced with a zero value. When the value of the input signal is in the range of th1 to th2, -th1 to -th2 (th1 <th2), the output signal gradually changes from 0. When the input signal is greater than or equal to th2 or less than or equal to -th2, the output signal is the same as the input signal.

  The minute signal extraction unit 1205 is configured to extract a component equal to or lower than the threshold value from the input signal, and thereby extracts a signal component having a minute amplitude value from the high-frequency signal component that is the filter input signal. FIG. 10B shows an example of input / output characteristics of the minute signal extraction unit 1205. As illustrated in FIG. 10B, the minute signal extraction unit 1205 outputs the same value as the input signal when the input signal is −th1 to th1. Further, the output signal gradually approaches 0 when the input signal is in the range of th1 to th2, -th1 to -th2 (th1 <th2). Furthermore, if the input signal is greater than or equal to th2 and less than or equal to -th2, the output value is zero.

  The operation of the modification shown in FIG. 9 will be described. When the high frequency signal component separated by the wavelet transform unit 110 is input, the micro signal extraction unit 1205 extracts a signal component having a micro amplitude value in the high frequency signal component. The IIR filter 1200 performs time integration processing on the signal component of the minute amplitude value extracted by the minute signal extraction unit 1205, and removes noise from the signal component of the minute amplitude value. Further, when the high frequency signal component separated by the wavelet transform unit 110 is input, the coring filter 1204 performs a coring process in which the signal component of the minute amplitude value in the high frequency signal component is set to 0 value and removed. Then, the signal after the time integration processing by the IIR filter 1200 and the signal after the processing by the coring filter 1204 are added by the adder 1203b and output as a high-frequency signal component after noise removal.

  According to the modification, the time integration filter unit performs the time integration process on the signal component of the minute amplitude value in the filter input signal. As a result, since the time integration process is not applied to signal components other than the signal component having a minute amplitude value, it is possible to prevent blurring from occurring at the edge portion of the moving object.

(Second Embodiment)
The image processing apparatus according to the second embodiment of the present invention will be described below with reference to the drawings. FIG. 11 shows the configuration of an image processing apparatus 2000 according to the second embodiment of the present invention.

  In FIG. 11, the image processing apparatus 2000 includes a wavelet transform unit 110, a wavelet inverse transform unit 500, and a motion detection unit 600. Similar to FIG. 1, the number of times of separation into subbands is two.

  11 is different from FIG. 1 in that a motion detection unit 600 is added and the configuration of the wavelet inverse transform unit 500 is different. The wavelet inverse transform unit 500 is an example of a subband inverse transform unit that synthesizes subband image signals, and the wavelet inverse transform is an example of a subband inverse transform.

  As shown in FIG. 11, the wavelet inverse transform unit 500 includes an edge signal creation unit 510, a time integration filter unit 120a, and an edge signal addition unit 520. The wavelet inverse transform unit 500 creates an edge signal, which will be described later, based only on the high frequency signal component transformed by the wavelet transform unit 110, removes noise in the time direction contained in the edge signal, and regenerates the image signal. And output as an output image signal. The motion detection unit 600 detects the motion of the input image signal. For example, the motion detection unit 600 detects the motion of the input image signal by detecting the difference between the input image signals.

  The edge signal creation unit 510 creates an edge signal obtained by performing wavelet inverse transform only on the high-frequency signal component among the high-frequency signal component and the low-frequency signal component converted by the wavelet transform unit 110. The edge signal creation unit 510 includes a level 1 edge signal creation unit 511 and a level 2 edge signal creation unit 512. Specific functions and operations of the level 1 edge signal creation unit 511 and the level 2 edge signal creation unit 512 will be described later.

  The time integration filter unit 120a removes noise in the time axis direction from the edge signal by performing time integration processing on the edge signal generated by the edge signal generation unit 510. Here, unlike the time integration type filter unit 120 shown in FIG. 1, the time integration type filter unit 120a is configured by one IIR filter. Specifically, the time integration type filter unit 120a is configured by one IIR filter 1200 of FIG. Further, the time integration type filter unit 120a may have the configuration of the modified example of FIG. However, in FIGS. 4 and 9, the input is a high-frequency signal component and the output is a high-frequency signal component after noise removal, whereas in the case of the time integration type filter unit 120a, the edge signal is processed as a high-frequency signal component. And is subject to noise removal. That is, the input is an edge signal input from the edge signal creation unit 510, and the output is an edge signal after noise removal. The specific operation is the same as that described with reference to FIGS.

  As shown in FIG. 11, the motion detection result is input from the motion detection unit 600 to the time integration type filter unit 120a. The time integration type filter unit 120a is configured to be able to change the time constant of the time integration process, and reduces the time constant k at the position where the motion is detected by the motion detection unit 600 to remove noise in the time axis direction. Perform the process. By reducing the time constant k in this manner, moving objects can be prevented from blurring.

  The value of the time constant k is set in advance and stored in a storage unit (memory) of the image processing apparatus, and may be read by the time integration type filter unit 120a and used for time constant control. Two time constants k are set corresponding to each of the stationary position and the position where the motion is detected, and the time constant k may be switched depending on the presence or absence of the motion. Further, the time constant may be changed according to the degree of movement. In this case, the time constant k may be expressed as a function of the amount of motion or may be set in the form of a map.

  The edge signal addition unit 520 adds the edge signal subjected to the time integration processing by the time integration type filter unit 120a and the low-frequency signal component converted by the wavelet conversion unit 110, thereby regenerating the image signal from which noise has been removed. This is configured and output as an output image signal.

  Next, the operation of the image processing apparatus 2000 according to the second embodiment of the present invention will be described with reference to the drawings. Also in the present embodiment, an example of performing wavelet transform based on the Haar base will be described. The number of times of conversion into subbands is two.

First, the wavelet transform unit 110 performs wavelet transform on an input image signal, thereby separating the input image signal into a low-frequency signal component and a high-frequency signal component. The procedure of wavelet transform by the wavelet transform unit 110 is the same as that of the first embodiment, and is as described with reference to FIGS. By the conversion by the wavelet transform unit 110, the input image signal is separated into frequency band components of level 1 converted signal ₁ LL _0,0 , ₁ HL _0,0 , ₁ LH _0,0 and ₁ HH _0,0 , Further, the low-frequency signal component ₁ LL of level ₁ is separated into components of each frequency band of level 2 conversion signal ₂ LL _0,0 , ₂ HL _0,0 , ₂ LH _0,0 and ₂ HH _0,0. . As a result, the input image signal includes a low-frequency signal component ₂ LL _0,0 and a high-frequency signal component ₁ HL _0,0 , ₁ LH _0,0 , ₁ HH _0,0 , ₂ HL _0,0 , ₂ LH _{0. , 0} and ₂ HH _0,0 . Note that these are calculated according to the above equation (1).

Next, the edge signal creation unit 510 creates an edge signal by performing wavelet inverse transform only on the high-frequency signal component transformed by the wavelet transform unit 110. Here, the edge signal of level n can be defined as shown in the following equation (3).
_{n EL 2V, 2H = (n} HL V, H + n LH V, H + n HH V, H) / 4
_n EL _{2V, 2H + 1} = (− _n HLV _{, H} + _n LHV _{, H} − _n HHV _{, H} ) / 4
_{n EL 2V + 1,2H = (n} HL V, H - n LH V, H - n HH V, H) / 4
_{n EL 2V + 1,2H + 1 =} (- n HL V, H - n LH V, H + n HH V, H) / 4
... (3)

  Comparing equation (2) and equation (3), equation (3) does not have the low-frequency signal component of equation (2) (it is 0). Accordingly, the wavelet inverse transform is performed only on the high frequency signal component by excluding the low frequency signal component according to the equation (3). This conversion result is an image signal obtained by synthesizing high-frequency signal components, and therefore has information on the edge portion. Therefore, in the present invention, as described above, this conversion result signal is defined as an edge signal.

The following equation (4) is derived from the equations (2) and (3).
_n LL _{2V, 2H} = _{n + 1} LL _{V, H} / 4 + _{n + 1} EL _{2V, 2H
n} LL _{2V, 2H + 1} = _{n + 1} LL _{V, H} / 4 + _{n + 1} EL _{2V, 2H + 1
n} LL _{2V + 1,2H} = _{n + 1} LL _{V, H} / 4 + _{n + 1} EL _{2V + 1,2H
n} LL _{2V + 1,2H + 1} = _{n + 1} LL _{V, H} / 4 + _{n + 1} EL _{2V + 1,2H + 1}
... (4)

  As shown in Expression (4), the n-level low-frequency signal component can be represented by the (n + 1) -level edge signal and the low-frequency signal component. Therefore, if there is an edge signal component and a low-frequency signal component at the highest level, the low-frequency signal component at the lowest level 0 can be restored.

For example, when the wavelet transform is performed up to level 2, the signal component ₀ LL _1,1 of level 0 at the horizontal position 1 and the vertical position 1 is expressed by the following equation (5).
₀ LL _1,1 = ₁ LL _0,0 / 4 + ₁ EL _1,1
= ( ₂ LL _0,0 / 4 + ₂ EL _0,0 ) / 4 + ₁ EL _1,1
= ₂ LL _0,0 / 16 + ₂ EL _0,0 / 4 + ₁ EL _1,1 (5)

Therefore, the level 2 edge signal generation unit 512 of the edge signal generation unit 510 performs level 2 edge signal ₂ from ₂ HL _0,0 , ₂ LH _0,0 and ₂ HH _0,0 converted by the level 2 conversion unit 122. Create EL _1,1 . The level 1 edge signal creation unit 511 creates the level 1 edge signal ₁ EL _1,1 from ₁ HL _0,0 , ₁ LH _0,0 and ₁ HH _0,0 converted by the level 1 conversion unit 121. Then, the edge signal generating unit 510, adds a quarter of ₂ EL _{1, 1} created, and ₂ EL _{1, 1} created by the level 1 edge signal generator 511 by the level 2 edge signal generator 512 Then, an edge signal is created. The edge signal is as shown in Equation (6).
Edge signal = ₂ EL _0,0 / 4 + ₁ EL _1,1 (6)

  By performing such processing for all positions, the edge signal creation unit 510 calculates an edge signal by performing wavelet inverse transform only on the high-frequency signal component obtained by the wavelet transform unit 110.

The time integration type filter unit 120a removes noise in the time axis direction from the edge signal by performing time integration processing on the edge signal generated by the edge signal generation unit 510. Most of the noise in the input image signal is contained in the high frequency signal component as compared with the low frequency signal component. The edge signal is calculated by adding the high frequency signal component, and therefore, the noise included in the input image signal is not included in the low frequency signal component ₂ LL _0,0 , but is included in the edge signal. Therefore, by performing the time integration process on the edge signal by the time integration filter unit 120a, the noise of the input image signal can be effectively removed.

  Here, the edge signal is obtained by performing wavelet inverse transformation only on the high-frequency signal component. The edge signal is obtained by adding the high-frequency signal component as described using Expression (3). In this embodiment, time integration processing is performed using such an edge signal as a high-frequency signal component. Therefore, one IIR filter 1200 in the time integration filter 120a can be configured, and the frame memory can be saved. be able to. As a result, the configuration of the time integration filter can be simplified.

  Further, the time integration type filter unit 120a changes the time constant k according to the movement of the input image signal as follows. The motion detection unit 600 detects the motion of the input image signal and inputs the motion detection result to the time integration type filter unit 120a. Then, the time integration type filter unit 120a performs a process of reducing noise in the time axis direction by reducing the time constant k at the position where the motion is detected by the motion detection unit 600. By reducing the time constant k in this manner, moving objects can be prevented from blurring.

The edge signal adding unit 520 adds the edge signal subjected to the time integration processing by the time integrating filter 120a and the low-frequency signal component converted by the level 2 conversion unit 112, thereby removing the output image signal from which noise has been removed. Is calculated. Here, the low-frequency signal component ₀ LL ′ _1,1 of level 0 is calculated by adding the edge signal and 1/16 of the low-frequency signal component ₂ LL _0,0 according to equation (5).

  According to the image processing apparatus of the second embodiment of the present invention as described above, an edge signal obtained by performing subband inverse transform only on a high-frequency signal component is created, and the edge signal is converted into a high-frequency signal. Time integration processing is performed as a component. Therefore, the high-frequency signal component can be processed by one time integration filter, and the configuration can be simplified.

  Unlike the conventional general wavelet inverse transform, in this embodiment, the wavelet inverse transform process is performed in two stages. The first stage process is an edge signal creation process, which creates an edge signal by performing inverse transformation only for the high frequency signal component. The edge signal can be said to be an intermediate conversion signal. The second stage processing is edge signal addition processing, in which the edge signal and the low frequency signal are added. Thereby, the final inverse transform signal is calculated, and the inverse transform process is completed. In this embodiment, the time integration process is performed between the first stage process and the second stage process. That is, the intermediate stage inverse transform signal obtained by the first stage process becomes the target of the time integration process. As a result, only one IIR filter is required, the frame memory is saved, and the configuration is simplified.

  In the image processing apparatus according to the second embodiment of the present invention, the time integration type filter unit is configured to be able to change the time constant of the time integration process. Thereby, the effect of time integration processing can be adjusted and noise removal can be suitably performed.

  Further, according to the image processing apparatus of the second embodiment of the present invention, the motion detection unit that detects the motion of the input image signal is provided, and the time integration type filter unit is a position where the motion is detected by the motion detection unit. The time constant is changed. Thereby, it is possible to prevent blurring from occurring at the edge portion of the moving object.

  In the image processing apparatus according to the second embodiment of the present invention, the motion detection unit has a configuration for detecting motion by detecting a difference between input image signals. Thereby, the motion of the input image signal can be suitably detected.

(Third embodiment)
Hereinafter, an image processing apparatus according to a third embodiment of the present invention will be described with reference to the drawings. FIG. 12 is a block diagram illustrating a configuration of an image processing device 3000 according to the third embodiment.

  In FIG. 12, an image processing device 3000 includes a wavelet transform unit 110, a coring process unit 140, a wavelet inverse transform unit 130a, a minute signal extraction unit 150, a minute edge signal creation unit 160, and a time integration filter unit. 120b and an edge signal adder 170. The number of separation into subbands is two. The wavelet transform unit 110, the coring processing unit 140, and the wavelet inverse transform unit 130a in FIG. 12 correspond to the wavelet transform unit 710, the coring process unit 740, and the wavelet inverse transform unit 730 in FIG.

  The wavelet transform unit 110 performs a wavelet transform on the input image signal, thereby separating the frequency band into a low frequency band including a low frequency signal component and one or more high frequency bands including a high frequency signal component. Each sub-band image signal is converted. The wavelet transform unit 110 is an example of a subband transform unit of the present invention, and the wavelet transform is an example of a subband transform. The wavelet transform unit 110 includes a level 1 transform unit 111 and a level 2 transform unit 112. The level 1 converter 111 performs the first (level 1) wavelet transform on the input image signal, and the level 2 converter 112 performs the second (level 2) wavelet on the image signal after the first wavelet transform. Perform conversion.

  The coring processing unit 140 performs a coring process for removing a signal component having a minute amplitude value from the high frequency signal component obtained by the wavelet transform unit 110. As shown in FIG. 12, the coring processing unit 140 includes a first high frequency signal coring unit 141 and a second high frequency signal coring unit 142. The first high frequency signal coring unit 141 processes the high frequency signal component obtained by the level 1 conversion unit 111, and the second high frequency signal coring unit 142 is the high frequency signal component obtained by the level 2 conversion unit 112. Process.

  The wavelet inverse transform unit 130a performs wavelet inverse transform on the high-frequency signal component subjected to coring processing by the coring processing unit 140 and the low-frequency signal component obtained by the wavelet transform unit 110, thereby re-generating the image signal. Constitute. The wavelet inverse transform unit 130a is an example of a subband inverse transform unit that synthesizes subband image signals, and the wavelet inverse transform is an example of a subband inverse transform.

  The wavelet inverse transform unit 130a includes a level 1 inverse transform unit 131a and a level 2 inverse transform unit 132a. The level 2 inverse transform unit 132a performs wavelet inverse transform on the high frequency signal component subjected to the coring process by the second high frequency signal coring unit 142 and the low frequency signal component obtained by the wavelet transform unit 110. Do. The level 1 inverse transform unit 131a performs wavelet inverse transform on the signal component obtained by the level 2 inverse transform unit 132a and the high frequency signal component subjected to coring processing by the first high frequency signal coring unit 141. Do. Thereby, the separated high-frequency signal component and low-frequency signal component are synthesized, and the image signal is reconstructed.

  The minute signal extracting unit 150 extracts a signal component having a minute amplitude value from the high frequency signal component converted by the wavelet transform unit 110. As shown in FIG. 12, the minute signal extraction unit 150 includes a first minute signal extraction unit 151 and a second minute signal extraction unit 152. The first minute signal extraction unit 151 extracts a signal component having a minute amplitude value from the level 1 high-frequency signal component obtained by the level 1 conversion unit 111. The second minute signal extraction unit 152 extracts a signal component having a minute amplitude value from the level 2 high-frequency signal component obtained by the level 2 conversion unit 112.

  The minute edge signal creation unit 160 creates an edge signal with a minute amplitude value by performing wavelet inverse transformation on the high frequency signal component of the minute amplitude value extracted by the minute signal extraction unit 150.

  As shown in FIG. 12, the minute edge signal creation unit 160 includes a level 1 minute edge signal creation unit 161 and a level 2 minute edge signal creation unit 162. The level 2 minute edge signal creation unit 162 performs wavelet inverse transform on the high frequency signal component of the minute amplitude value extracted by the second minute signal extraction unit 152, thereby converting the level 2 edge signal of the minute amplitude value. create. The level 1 minute edge signal creation unit 161 performs wavelet inverse transform on the high frequency signal component of the minute amplitude value extracted by the first minute signal extraction unit 151, thereby converting the level 1 edge signal of the minute amplitude value. create. Further, the minute edge signal creation unit 160 adds the edge signals of level 1 and level 2 of the minute amplitude value to create an edge signal of the minute amplitude value.

  A specific creation method of the edge signal having a minute amplitude value is created by the same processing as the creation method of the edge signal in the second embodiment. However, in the second embodiment, the entire high-frequency signal component is processed, whereas in the present embodiment, the signal component of a minute amplitude value in the high-frequency signal component is processed. A signal component having a minute amplitude value is obtained.

  The time integration type filter unit 120b performs time integration processing on the edge signal having a minute amplitude value created by the minute edge signal creating unit 160, thereby removing noise in the time axis direction from the edge signal having the minute amplitude value. . Here, the configuration of the time integration type filter unit 120b may be the same as that of the time integration type filter 120a of FIG. 11, and includes only one filter configuration shown in FIG. 4 or FIG. However, in the case of the time integration type filter unit 120b, the input is an edge signal having a minute amplitude value, and the output is an edge signal having a minute amplitude value from which noise has been removed. The specific operation is the same as that described with reference to FIGS.

  The edge signal adding unit 170 adds the edge signal having a minute amplitude value subjected to the time integration processing by the time integrating filter unit 120b and the image signal obtained by the wavelet inverse transform unit 130a, thereby removing the noise-removed image signal. Is reconstructed and output as an output image signal.

  Next, the operation of the image processing device 3000 according to the third embodiment of the present invention will be described with reference to the drawings. Also in the present embodiment, an example of performing wavelet transform based on the Haar base will be described.

  First, the wavelet transform unit 110 separates an input image signal into a low-frequency signal component and a high-frequency signal component by performing wavelet transform. The high-frequency signal component is input to the coring processing unit 140 and the minute signal extraction unit 150, and the low-frequency signal component is input to the wavelet inverse transformation unit 130a.

  The coring processing unit 140 removes a signal component having a minute amplitude value from the input high frequency signal component. The high-frequency signal component after the coring process is input to the wavelet inverse transform unit 130a. The wavelet inverse transform unit 130a performs wavelet inverse transform on the high frequency signal component input from the coring processing unit 140 and the low frequency signal component input from the wavelet transform unit 110, thereby generating an image signal. The generated image signal is input to the edge signal adding unit 170.

  The minute signal extraction unit 150 extracts a signal component having a minute amplitude value in the input high frequency signal component for each level. The high frequency signal component of the minute amplitude value extracted by the minute signal extraction unit 150 is input to the minute edge signal creation unit 160. The minute edge signal creation unit 160 performs wavelet inverse transformation on the input high frequency signal component of the minute amplitude value, and creates an edge signal of the minute amplitude value. The edge signal having a minute amplitude value is input to the time integration type filter unit 120b.

  The time integration type filter unit 120b performs time integration processing on the input edge signal having a minute amplitude value, and removes noise in the time axis direction from the edge signal having the minute amplitude value. The edge signal having a minute amplitude value subjected to the time integration process is input to the edge signal adding unit 170. The edge addition unit 170 adds the edge signal having a small amplitude value processed by the time integration filter unit 120b and the image signal input from the wavelet inverse conversion unit 130a, and outputs an output image signal from which noise has been removed. To do.

    According to the image processing apparatus of the third embodiment of the present invention as described above, a signal component having a minute amplitude value in the high frequency signal component is extracted, and the subband inverse is performed on the high frequency signal component having the minute amplitude value. An edge signal having a minute amplitude value obtained by performing the conversion is created, and time integration processing is performed using the edge signal having the minute amplitude value as a high-frequency signal component. As a result, since the time integration process is not applied to signal components other than the signal component having a minute amplitude value, it is possible to prevent blurring from occurring at the edge portion of the moving object, and to convert the edge signal into time. Since the integration process is performed, the high-frequency signal component can be processed by one time integration filter, and the configuration can be simplified.

  The embodiments of the present invention have been described above by way of example, but the scope of the present invention is not limited to these embodiments, and can be changed or modified according to the purpose within the scope of the claims. is there.

  In the above description, the motion detection unit 600 is provided in the image processing apparatus 2000 according to the second embodiment. Similar motion detection units may be provided in the image processing apparatuses 1000 and 3000 according to the first and third embodiments. The time integration type filter unit may perform a process of reducing noise in the time axis direction by reducing the time constant k at a position where the motion is detected by the motion detection unit. As a result, the moving object can be prevented from blurring as described above.

  Further, in the configuration of the time integration type filter unit in the first to third embodiments, the influence of the past signal appears due to the time constant k of the IIR filter. Therefore, the image processing apparatus may have a configuration for performing a coring process for removing a signal component having a minute amplitude value when outputting a signal from which noise has been removed (filtered signal). Thereby, the influence of the past noise can be reduced.

  As described above, the present invention eliminates the noise in the time axis direction from the high frequency signal component by performing the time integration process on the high frequency signal component, thereby preventing the edge portion in the image from blurring. It has the effect of reducing noise of the entire image including the peripheral portion of the edge while preventing it, a camera monitor mounted in a car, a video camera as an imaging device, a television having an image display function, and an image recording function It is useful for an image processing apparatus used for a DVD recorder or the like having

1 is a block diagram showing a configuration of an image processing apparatus according to a first embodiment of the present invention. (A) The figure which showed the acquisition position of each sample of an input image signal (b) The figure which shows the subband separation state after converting the input image signal into the subband to level 2 (A) Diagram showing procedure of wavelet transform (b) Diagram showing procedure of wavelet transform (c) Diagram showing procedure of wavelet transform The figure which shows the structure of the IIR filter as a specific example of a time integration type filter part (A) The figure which shows the image signal before noise removal (b) The figure which shows the high frequency signal component before noise removal (A) The figure which shows the high-frequency signal component of the time-axis direction before noise removal (b) The figure which shows the high-frequency signal component of the time-axis direction after noise removal (c) The figure which shows the high-frequency signal component after noise removal (A) The figure which shows the procedure of wavelet inverse transformation (b) The figure which shows the procedure of wavelet inverse transformation (c) The figure which shows the procedure of wavelet inverse transformation (A) The figure which shows the output image signal from which the noise after wavelet inverse transformation was removed (b) The figure which shows the output image from which the noise after wavelet inverse transformation was removed The figure which shows the structure of the modification of a time integration type filter part (A) The figure which shows the example of the input / output characteristic of a time integration type filter part (b) The figure which shows the example of the input / output characteristic of a time integration type filter part The block diagram which shows the image processing apparatus in the 2nd Embodiment of this invention The block diagram which shows the image processing apparatus in the 3rd Embodiment of this invention Block diagram showing the configuration of a conventional image processing apparatus (A) Diagram showing an example of an input image (b) Diagram showing an example of an input image signal (c) Diagram showing a high-frequency signal component of the input image signal (A) The figure which shows the high frequency signal component of the input image signal after noise removal by the conventional image processing apparatus (b) The figure which shows an example of the output image of the conventional image processing apparatus

Explanation of symbols

1000, 2000, 3000 Image processing device 110 Wavelet transform unit 120, 120a, 120b Time integration type filter unit 130, 130a Wavelet inverse transform unit 140 Coring processing unit 150 Minute signal extraction unit 160 Minute edge signal creation unit 170 Edge signal addition unit 500 Wavelet inverse transform unit 510 Edge signal creation unit 520 Edge signal addition unit 600 Motion detection unit 1200 IIR filter 1201 Frame delay memory 1202a, 1202b Multiplier 1203, 1203b Adder 1204 Coring filter 1205 Minute signal extraction unit

Claims (9)

A subband conversion unit that separates the input image signal into a high-frequency signal component and a low-frequency signal component by performing subband conversion on the input image signal;
An image processing apparatus comprising: a time integration type filter unit that removes noise in the time axis direction from the high frequency signal component by performing time integration processing on the high frequency signal component. An edge signal creation unit that creates an edge signal obtained by performing subband inverse transform only on the high-frequency signal component among the high-frequency signal component and the low-frequency signal component;
The time integration filter unit performs time integration processing with the edge signal as the high-frequency signal component,
The image processing apparatus according to claim 1, further comprising an edge signal addition unit that adds the edge signal processed by the time integration filter unit and the low-frequency signal component.   The time integration type filter unit performs time integration processing on the signal component of the minute amplitude value in the filter input signal, and removes the signal after time integration processing and the signal component of the minute amplitude value in the filter input signal. The image processing apparatus according to claim 1, wherein a signal subjected to the coring process is added. A coring processing unit for removing a signal component of a minute amplitude value in the high-frequency signal component;
A subband inverse transform unit that performs subband inverse transform on the high frequency signal component obtained by the coring processing unit and the low frequency signal component;
A minute signal extraction unit for extracting a signal component of a minute amplitude value in the high frequency signal component;
A minute edge signal creating unit that creates an edge signal of a minute amplitude value obtained by performing subband inverse transformation on the high frequency signal component of the minute amplitude value extracted by the minute signal extracting unit;
The time integration type filter unit performs time integration processing using the edge signal of the minute amplitude value as the high-frequency signal component,
Further, an edge signal addition unit is provided for adding the edge signal having the minute amplitude value processed by the time integration type filter unit and the signal obtained by the subband inverse transform unit. Item 8. The image processing apparatus according to Item 1.   5. The time integration filter unit includes a cyclic filter that adds a filter input signal and a filtered signal of a previous frame at a predetermined ratio. Image processing apparatus.   The image processing apparatus according to claim 5, wherein the time integration filter unit performs a coring process for removing a signal component having a minute amplitude value in a signal processed by the cyclic filter.   The image processing apparatus according to claim 1, wherein the time integration type filter unit is configured to be able to change a time constant of time integration processing. A motion detector for detecting the motion of the input image signal;
The image processing apparatus according to claim 7, wherein the time integration filter unit changes the time constant at a position where the motion is detected by the motion detection unit.   The image processing apparatus according to claim 8, wherein the motion detection unit detects the motion by detecting a difference between the input image signals.