JP2011055415A - Device and method for image processing - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To effectively remove noise components of a specific frequency which is present in a plurality of frame images. <P>SOLUTION: The device for image processing includes: an image input portion 1 for inputting each frame image composing a moving picture successively along time; a frequency resolution portion 2 for dividing the moving picture from the image input portion 1 into a plurality of frequency bands; a coefficient changing portion 3 for erasing or reducing components of the divided moving picture which specially depend on time among the respective divided moving pictures; an image recomposing portion 4 which recomposes all of the moving pictures divided in the frequency resolution portion 2 by performing processing reverse to that of the frequency resolution portion 2 based on the erased or reduced components in the coefficient changing portion 3, divides the recomposed moving picture for each time, and creates the restored frame image; and an image output portion 5 for outputting the frame image restored in the image recomposing portion 4. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to an image processing apparatus and an image processing method capable of sharpening various images such as an ultrasonic image.

  Currently, in the medical field, many medical images such as an X-ray image, an MRI image, and a CT image are used for diagnosis taking advantage of the feature that it is very easy to understand visually.

  Among them, unlike X-ray images, ultrasound images are less exposed to exposure during imaging, and the imaging device is cheaper and smaller than MRI images, enabling real-time diagnosis. It is used for diagnosis in various fields including the field and obstetrics and gynecology. However, ultrasonic images are more affected by noise than other medical images, and tissue boundary images are often not obtained clearly, and inexperienced examiners judge when the lesion is small. It becomes difficult and skill is required to make an accurate diagnosis.

  In recent years, there has been an increasing demand for a low-priced ultrasonic imaging apparatus that can be used by anyone, not just medical professionals, for the purpose of healthcare, for example, for the treatment of metabolic syndrome. In that case, since it is important to obtain an easy-to-understand image that can be accurately diagnosed by ordinary people, there is a need for an image processing method and an image processing apparatus capable of reducing image noise and clarifying tissue boundaries. It was.

  For such a problem, in Patent Document 1, for each frame image constituting a moving image, the first frame image is decomposed into a plurality of frequency regions, and coefficients belonging to a desired frequency band among the plurality of frequency regions, After changing using the coefficients belonging to the same frequency domain of the second frame image, which is the frame image before the first frame image, and finishing the coefficient changing process for all frequency domains, the image is based on the changed coefficient. An image processing method has been proposed in which image signals can be enhanced while suppressing unnecessary noise enhancement by reconstructing the image and outputting the image as a first frame image.

JP 2007-94555 A

  In the above-mentioned Patent Document 1, a proposal has been made to clarify an image in consideration of a difference in analysis results of two-dimensional discrete wavelet transform performed on adjacent frame images. It is difficult to filter a noise component having a low target frequency, and further image sharpening has been demanded.

  The present invention has been made in view of the above problems, and an object thereof is to provide an image processing apparatus and an image processing method capable of effectively removing a noise component having a specific frequency present in a plurality of frame images.

  In order to achieve the above object, an image processing apparatus according to the present invention includes an input unit that sequentially inputs each frame image constituting a moving image, a decomposing unit that divides the moving image into a plurality of frequency bands, and the divided Among the moving images, among the components of the divided moving images depending on time, changing means for deleting or reducing the expansion coefficient of the target frequency component, and the decomposing means based on the deleted or reduced components Reconstructing means for reconstructing all the divided moving images by dividing the reconstructed moving images every time by performing processing reverse to the processing, and the restoration Output means for outputting the frame image thus obtained.

  The changing means here has a configuration for enlarging the divided moving image components showing high-frequency components that do not depend on time in the divided moving images.

  The decomposition means performs wavelet transform processing in each of the vertical, horizontal, and time directions of the moving image, and the reconstructing means performs the vertical, horizontal, and time of all the divided moving images. A wavelet inverse transform process is performed for each of the directions.

  In this case, the component is a wavelet expansion coefficient obtained by performing the wavelet transform process.

  Further, the wavelet transform process by the decomposing means and the wavelet inverse transform process by the reconstructing means are preferably performed using a Daubecheies 10 mother wavelet.

  In order to achieve the above object, the image processing method of the present invention includes an input step for sequentially inputting each frame image constituting a moving image, a decomposition step for dividing the moving image into a plurality of frequency bands, and the division. Among the moving images, among the components of the divided moving images depending on time, a changing step of deleting or reducing the expansion coefficient of the target frequency component, and the decomposing means based on the deleted or reduced components Reconstructing all the divided moving images by performing processing reverse to the above, dividing the reconstructed moving images every time, and generating a restored frame image; and the restoration An output step for outputting the frame image.

  The same object can be achieved for a program that causes a computer to execute such an image processing method.

  Further, the same object can be achieved for a computer-readable storage medium storing the program.

  In the image processing apparatus and the image processing method, each frame image constituting the moving image is not decomposed into a plurality of frequency bands, but the three-dimensional moving image itself is decomposed into a plurality of frequency bands. For a divided moving image depending on time, an arbitrary frequency component is arbitrarily selected, deleted or reduced, and all divided moving images are reconstructed, so that they exist in a plurality of frame images. It becomes possible to effectively remove a noise component of a specific frequency.

  This can also be realized for a program that causes a computer to execute the image processing method and a computer-readable storage medium that stores the program.

  In addition, by enlarging the divided moving image components showing high-frequency components that do not depend on time, and reconstructing all the divided moving images, a sharper image with enhanced edges for each restored frame image Can be obtained.

  Furthermore, with regard to the processing performed by the decomposing means and the reconstruction means, it is possible to use each of the well-known wavelet transform and inverse wavelet transform processes, and then change the wavelet expansion coefficient to make a clearer frame. An image can be output.

  In addition, by using the Daubecheies 10 mother wavelet, it is possible to realize excellent image reconstruction and obtain good resolution in obtaining a final frame image.

1 is a block configuration diagram of an image processing apparatus showing a preferred embodiment of the present invention. It is a block diagram which shows the structure of a frequency decomposition part same as the above. FIG. 3 is a plan view of an image schematically showing band division of a two-dimensional signal. It is the schematic which shows the procedure of the band division | segmentation of a two-dimensional signal same as the above. It is the schematic explaining a series of processing procedures by 3D wavelet analysis same as the above. It is a block diagram which shows the structure of an image reconstruction part same as the above. It is a schematic block diagram of an ultrasonic echo image system containing the image processing apparatus of FIG. 1 same as the above. FIG. 8 is a plan view of each image obtained by the system of FIG. FIG. 9 is an enlarged view of the main part of FIG.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

  FIG. 1 shows a functional configuration of an image processing apparatus 100 in the present embodiment. In FIG. 1, an image processing apparatus 100 includes an image input unit 1, a frequency resolving unit 2, a coefficient changing unit 3, an image reconstruction unit 4, and an image output unit 5.

  The image input unit 1 inputs an image (image signal) of each frame constituting a moving image. As an image supply source, for example, although not shown, there is an imaging device incorporated in a medical diagnostic imaging apparatus. An ultrasonic image, an X-ray image, an MRI image, a CT image, and the like from this imaging apparatus are sent to the image input unit 1 as an electrical analog image signal. The image input unit 1 performs A / D conversion on the analog image signal, and sends the analog image signal to the subsequent frequency resolution unit 2 as a digital image signal formed by two-dimensionally arranging a plurality of pixels.

  In the present embodiment, it is assumed that the moving image input to the image input unit 1 has a predetermined frame rate and each frame image has a predetermined resolution.

  In addition, the image of each frame may be a monochrome image composed of a single color component or a color image composed of a plurality of color components. In the case of the latter color image, each process described later is performed on each color component.

  When the image of each frame has a plurality of color components, the image input unit 1 may include a color space conversion unit 11. The color space conversion unit 11 converts the color space from RGB to YUV when the image of each frame input to the image input unit 1 has three color components of RGB. For the subsequent processing, the color space conversion unit 11 What is necessary is just to make it with respect to each color component after converting in the conversion part 11. FIG.

  An image is input to the frequency resolution unit 2 in units of frames from the image input unit 1, and the frequency resolution unit 2 here particularly uses a plurality of frame image groups arranged in time series as one moving image data. This has a function of decomposing this into a plurality of frequency bands and sending out the decomposed frequency band coefficients of the frame image group to the coefficient converting unit 3.

  As a technique for decomposing the plurality of frame image groups into a plurality of frequency bands, in the present embodiment, a function of performing multi resolution analysis (MRA) using a three-dimensional discrete wavelet transform is used. Provided. The frequency decomposing unit 2 can analyze the feature by repeatedly performing a processing routine for decomposing the frame image group into a low frequency component and a high frequency component a plurality of times.

  The functional configuration of the frequency resolving unit 2 will be described in more detail. As shown in FIG. 2, the frequency resolving unit 2 includes a plurality of low-pass filters (LPF) 61 to 63, a plurality of high-pass filters (HPF) 64 to 66, Each of the filters 61 to 66 is configured by a combination with down samplers 67 to 72 that reduce the frequency band of the signal component to 1 / N (1 / N). The configuration of FIG. 2 shows a configuration example when octave division, which is the most general wavelet transform process, is performed over a plurality of levels. The number of levels representing the number of layers is 3 (level 1 to level 3), and the image signal is divided into a low frequency and a high frequency, and only the low frequency components are hierarchically divided. The number of levels may be other than 3.

  In FIG. 2, the image signal input to the frequency resolving unit 2 is first band-divided by a level 1 low-pass filter 61 and a high-pass filter 64, and the obtained low-frequency signal and high-frequency signal correspond to each other. The downsamplers 67 and 68 reduce the frequency band (resolution) to 1 / N times.

  The low-frequency signal output from the down sampler 67 is further band-divided by the next level 2 low-pass filter 62 and high-pass filter 65. The frequency band (resolution) of the low-frequency side signal and the high-frequency side signal obtained by these band divisions is reduced to 1 / N times by the corresponding down samplers 69 and 70, respectively.

  The low-frequency signal output from the down sampler 69 is further band-divided by the next level 3 low-pass filter 63 and high-pass filter 66. The frequency band (resolution) of the low-frequency side signal and the high-frequency side signal obtained by these band divisions is reduced to 1 / N times by the corresponding downsamplers 71 and 72, respectively.

  By performing such processing up to a predetermined level, signals in each frequency band obtained by hierarchically dividing the low-frequency signal are sequentially generated. In the example of FIG. 2, it is shown that the LLL signal, the LLH signal, the LH signal 111, and the H signal are generated as a result of the band division to level 3. Here, “L” represents a low frequency component, and “H” represents a high frequency component.

FIG. 3 illustrates band components obtained as a result of band division of a two-dimensional original image up to level 3. However, the notation of “L” and “H” in FIG. 3 is different from the above description in which one-dimensional signals are handled. The symbol notation on the left represents the frequency component in the horizontal direction of the image, the symbol notation on the right represents the frequency component in the vertical direction of the image, and the leftmost subscript (subscript) represents the level. Therefore, for example, “HL ₁ ” indicates a high frequency component in the horizontal direction of the image at level 1 and a low frequency component in the vertical direction of the image. Also in this case, it is possible to extract a lower frequency band signal by repeatedly performing band division on the “LL” region at each level.

The concept of such band division is schematically shown in FIG. Here, the original original image to be subjected to level 0 band division is expressed as “LL ₀ ”, and from there, the level 1 low pass filter 61 and the high pass filter 64 are used for each of the horizontal and vertical directions of the image. Then, the divided images “LL ₁ ”, “HL ₁ ”, “LH ₁ ”, and “HH ₁ ” divided into the low frequency and the high frequency are obtained.

Each of these divided images is thinned out to 1 / N times by the down samplers 67 and 68, and then, for the divided image of “LL ₁ ”, a level 2 low-pass filter 62 and a high-pass filter 65 are obtained therefrom. Are used to obtain divided images of “LL ₂ ”, “HL ₂ ”, “LH ₂ ”, and “HH ₂ ” divided into a low frequency and a high frequency in each of the horizontal direction and the vertical direction of the image.

Each of these divided images is thinned out to 1 / N times by the down samplers 69 and 70, and then the level 3 low-pass filter 63 and the high-pass filter 66 are obtained from the divided image “LL ₂ ”. Are used to obtain divided images of “LL ₃ ”, “HL ₃ ”, “LH ₃ ”, and “HH ₃ ” divided into a low frequency and a high frequency in each of the horizontal direction and the vertical direction of the image.

Each of these divided images is thinned out to 1 / N times by the down samplers 71 and 72, and finally, "HL ₁ ", "LH ₁ ", "HH" as shown in FIG. ₁ , “HL ₂ ”, “LH ₂ ”, “HH ₂ ”, “LL ₃ ”, “HL ₃ ”, “LH ₃ ”, and “HH ₃ ” can be obtained.

  The configuration and method of the multi-resolution analysis using the wavelet transform relating to the one-dimensional signal and the two-dimensional signal have been described above. In the present invention, the frequency resolution unit 2 has the same multi-resolution degree particularly for three-dimensional moving image data. Has the function of performing analysis. As shown in FIG. 5, the frequency resolving unit 2 within the predetermined time for the image 110 of each frame sequentially sent out from the image input unit 1 along the time axis T and developed on the horizontal axis X and the vertical axis Y. This is achieved by handling a plurality of images 110 as a single moving image data 120 and dividing the moving image data 120 according to the above-described method. As described above, in the present embodiment, when handling moving images, the horizontal axis X and the vertical axis Y are obtained by performing the wavelet transform process considering the time axis T instead of every frame image 110 obtained at a single time. In addition to the time axis T, the moving image data 120 can be separated for the high frequency component and the low frequency component.

  The specific processing procedure will be described. As shown in FIG. 5A, the frequency resolving unit 2 first converts the original three-dimensional moving image data 120 to be processed into a low-pass filter 61 of level 1 and a high-pass. Band division is performed by the filter 64, and “LLL”, “HLL”, “LHL”, “L” divided into a low frequency and a high frequency in each direction of the time axis T, the vertical axis Y, and the horizontal axis X of the moving image data 120. Each divided moving image data 120 ′ of “HHL”, “LLH”, “HLH”, “LHH”, “HHH” is obtained. The obtained divided moving image data 120 ′ is thinned out to 1 / N times by the corresponding downsamplers 67 and 68.

  Here, the symbol notation on the left represents the frequency component in the time axis T direction of the moving image data 120, the symbol notation in the center represents the frequency component in the Y axis direction of the moving image data 120, and the symbol notation on the right represents the moving image. The frequency component regarding the horizontal-axis X direction of the data 120 is shown. In FIG. 5A, the notation “** H” is omitted for the frequency component in the horizontal axis X direction for convenience.

  As shown in FIG. 5B, the frequency resolving unit 2 uses the level 2 low-pass filter 62 and the high-pass filter 65 for the “LLL” divided moving image data 120 ′, and uses the divided moving image data 120. “LLL”, “HLL”, “LHL”, “HHL”, “LLH”, “HLH”, divided into a low frequency and a high frequency in each of the time axis T, the vertical axis Y, and the horizontal axis X Each divided moving image data 120 ″ of “LHH” and “HHH” is obtained.

  Each of the divided moving image data 120 ″ is thinned by a down sampler 69, 70 by a frequency of 1 / N, and then the level 3 low-pass filter 63 is obtained from the LLL divided image. And “LLL”, “HLL”, “LHL” divided into a low frequency and a high frequency for each direction of the time axis T, the vertical axis Y, and the horizontal axis X of the divided moving image data 120 ′ using the high-pass filter 66. ”,“ HHL ”,“ LLH ”,“ HLH ”,“ LHH ”,“ HHH ”divided moving image data 120 ′ ″.

  Each of these divided moving image data 120 ′ ″ is thinned out to 1 / N times by the down samplers 71 and 72, and finally, from one moving image data 120, FIG. A plurality of divided divided moving image data 120 ′, 120 ″, 120 ′ ″ as shown in FIG.

  The low-pass filters 61 to 63 and the high-pass filters 64 to 66 necessary for performing the band division of the moving image data 120 correspond to a scaling function in wavelet transform and a wavelet function (mother wavelet), respectively. Various functions used as the mother wavelet are known, but here, the Daubecheies function ψ (x) expressed by the following equation 1 is applied.

In Equation 1, k is an integer, q _k is a sequence, and φ (x) is a scaling function. This scaling function φ (x) satisfies the to-scale relationship with Daubecheies' mother wavelet ψ (x) shown in Equation 1, and can be expressed by the following Equation 2.

In Equation 2, _pk is a sequence. These numerical sequences p _k and q _k have values determined for each mother wavelet.

On the other hand, the sequence c _{0, k} shown in the following Equation 3 is determined by Fourier transforming the luminance signal or color difference signal of each pixel constituting the original moving image data 120.

In the above equation 3, f (x) represents the original signal. If the original signal f (x) and the sequence c _{0, k} are known _{, then} the components of the sequence c _{0, k} as a starting point are asymptotically determined in consideration of the following equations 4 and 5. If it is decomposed into _{j−1, k} and a sequence d _{j−1, k} , the original moving image data 120 can be decomposed for each frequency component.

  Here, j is an integer of 0 or less, and the absolute value of j-1 indicates the number of repetitions of the decomposition process, and corresponds to the above-mentioned level. L is an integer.

In the above equations 4 and 5, when there is a sequence c _{0, k} representing the original signal f (x), the sequence c _{j−1, k} and the sequence d _{j−1, k} are obtained therefrom, that is, the next level This means that the functions f (x) and g (x) of the low frequency component and the high frequency component are obtained. Here, the functions f (x) and g (x) are in a relationship satisfying the orthogonal condition, and the Daubecheies mother wavelet ψ (x) itself has a property as an orthogonal wavelet. Equations (4) and (5) can be said to be equations in which the digital low-pass filter a _k and the digital high-pass filter b _k are convolved with the inputs c _{j, k} , respectively. These numerical sequences a _k and b _k have values determined for each mother wavelet.

In the above description, the following relational expressions hold for each of the numerical sequences a _k , b _k , p _k , and q _k .

Furthermore, in the present embodiment, the sequence h _k of Equation 6 is expressed by the following relationship because it uses the Daubecheies 10 mother wavelet ψ (x), which is excellent in image reconstruction described later and has good resolution as a filter. The value is determined as follows.

The frequency resolving unit 2 performs the scaling coefficient s _k ^(j) and the wavelet expansion coefficient ω _k ^(j), which are parameters necessary for final image reconstruction, when performing the band division of the moving image data 120 described above. This is calculated and sent to the subsequent coefficient changing unit 3. Here, the scaling coefficient s _k ^(j) can be expressed as follows based on the scaling function φ (x).

  Equation 8 is as follows according to the to-scale relationship described above.

Therefore, the scaling coefficients s _k ⁽¹⁾ , s _k ⁽²⁾ , s _k ^(3), ..., Which are successively higher in level, can be obtained from the level 0 scaling coefficient s _k ⁽⁰⁾ .

On the other hand, the wavelet expansion coefficient ω _k ^(j) can be expressed as follows based on the mother wavelet ψ (x).

  The above formula 10 is as follows according to the two-scale relationship.

Therefore, the wavelet expansion coefficient ω _k ^{(j) at} each level can be obtained from the scaling coefficient s _k ^(j−1) at a level smaller than that.

Thus, when the scaling coefficient s _k ^(j) and the wavelet expansion coefficient ω _k ^(j) obtained by performing the wavelet analysis on the moving image data 120 are sent from the frequency resolving unit 2 to the coefficient changing unit 3, the coefficient The changing unit 3 arbitrarily selects an appropriate frequency component, deletes or reduces the wavelet expansion coefficient ω _k ^(j) related to the frequency component, and performs processing to remove or reduce image noise of the target frequency component. .

As shown in FIG. 5C, this depends on the time axis T among the divided moving image data 120 ′, 120 ″, 120 ′ ″ of each frequency component obtained by the frequency resolving unit 2. This is achieved by deleting or reducing the wavelet expansion coefficient ω _k ^(j) of the divided moving image data 120A (the shaded portion in the figure).

On the other hand, the coefficient changing unit 3 in the present embodiment obtains an image in which an edge is emphasized for each reconstructed frame image 130 to be described later, moving image data 120 ′, 120 ″ that does not depend on the time axis T. , 120 ″ ′, the function of enlarging or multiplying the wavelet expansion coefficient ω _k ^(j) of the divided moving image data 120B (the hatched portion in the figure) indicating the high frequency component is provided.

Thus, when the change of the wavelet expansion coefficient ω _k ^(j) indicating an appropriate frequency component is added by the coefficient changing unit 3, the changed wavelet expansion coefficient ω _k ^(j) and the scaling coefficient s _k ^(j) are included. The parameter information is sent to the image reconstruction unit 4.

  The image reconstruction unit 4 reconstructs the decoded frame image 130 shown in FIG. 5 by performing processing (wavelet inverse transformation processing) reverse to the image analysis processing by the frequency resolution unit 2. As shown in FIG. 6, a plurality of low-pass filters (LPF) 81-83, a plurality of high-pass filters (HPF) 84-86, and the frequency band of signal components input to the filters 81-86 are multiplied by N times. The up-samplers 87 to 92 and adders 93 to 95 are combined. In the figure, the number of levels representing the number of hierarchies is 3, but this can be changed as appropriate in accordance with the number of levels of the frequency resolving unit 2.

  In FIG. 6, among the band components output from the frequency resolving unit 2, the LLL signal and the LLH signal are up-sampled N times by the up-samplers 87 and 88, respectively. The LLL signal up-sampled by the up-sampler 87 is filtered by the low-pass filter 81, and the LLH signal up-sampled by the up-sampler 88 is filtered by the high-pass filter 84, and each signal is band-synthesized by the adder 93. To do. By this processing, the level 3 inverse transformation is completed.

  Similarly, if the above inverse transformation process is repeated up to level 1, the final frame image 130 after inverse transformation is obtained.

  That is, the output signal from the adder 93 is further upsampled to an N-fold frequency band by the upsampler 89, filtered by the low-pass filter 82, and sent to the adder 94. Further, the LH signal from the frequency resolving unit 2 is further upsampled to an N-fold frequency band by the upsampler 90, filtered by the high pass filter 85, and sent to the adder 94. The adder 94 band-synthesizes the signals from the low-pass filter 82 and the high-pass filter 85, whereby the level 2 inverse conversion is completed.

  The output signal from the adder 94 is further upsampled to a frequency band of N times by the upsampler 91, filtered by the low-pass filter 83, and sent to the adder 95. The H signal from the frequency resolving unit 2 is further upsampled to an N-fold frequency band by the upsampler 92, then filtered by the high pass filter 86 and sent to the adder 95. The adder 95 band-synthesizes the signals from the low-pass filter 83 and the high-pass filter 86, whereby the level 1 inverse conversion is completed.

  Although the configuration and method of wavelet inverse transformation relating to a one-dimensional signal has been described above, in the present invention, the image reconstruction unit 4 has a function of performing similar wavelet inverse transformation particularly for three-dimensional moving image data.

  The specific processing procedure will be described. In FIG. 5C, the image reconstructing unit 4 relates to the divided moving image data 120 ″ ′ at level 3 with respect to the time axis T, the vertical axis Y, and the horizontal axis X. The up-sampler 87 up-samples “LLL” divided moving image data 120 ′ ″ each having a low-frequency component in each direction, and filters the low-frequency filter 81. On the other hand, each of the other divided moving image data 120 ″ ′ of “HLL”, “LHL”, “HHL”, “LLH”, “HLH”, “LHH”, “HHH” is separated by another upsampler 88. Up-sample and filter this with a high-pass filter 84. Each data from the low-pass filter 81 and the high-pass filter 84 is band-combined by the adder 93, whereby the “LLL” divided moving image data 120 ″ in FIG. 5B is reconstructed.

  Similarly, the “LLL” divided moving image data 120 ″ output from the adder 93 is up-sampled by the up-sampler 89 and filtered by the low-pass filter 82. On the other hand, each of the other moving image data 120 ′ ′ of “HLL”, “LHL”, “HHL”, “LLH”, “HLH”, “LHH”, “HHH” is uploaded by another upsampler 90 This is sampled and filtered with a high-pass filter 85. Each data from the low-pass filter 82 and the high-pass filter 85 is band-synthesized by the adder 94, whereby the “LLL” divided moving image data 120 ′ in FIG.

  Further, the “LLL” divided moving image data 120 ′ output from the adder 94 is upsampled by the upsampler 91, and is filtered by the lowpass filter 83. On the other hand, each of the other divided moving image data 120 ′ of “HLL”, “LHL”, “HHL”, “LLH”, “HLH”, “LHH”, “HHH” is upsampled by another upsampler 92. This is filtered by a high pass filter 86. The original moving image data 120 can be reconstructed by band-combining the data from the low-pass filter 83 and the high-pass filter 86 by the adder 95.

  The image reconstruction unit 4 further divides the restored moving image data 120 into a two-dimensional frame image 130 along the time axis T, selects one or more of them, and sends them to the image output unit 5. The image output unit 5 outputs a reconstructed image obtained by the image reconstructing unit 4, that is, a frame image 130, which is realized by, for example, a display device or a printing device.

In the course of such a series of processing, the characteristics of the individual low-pass filters 81 to 83 and the high-pass filters 84 to 86 are determined by the wavelet expansion coefficient ω _k ^(j) and the scaling coefficient s _k ^(j) from the coefficient changing unit 3. Is done. That is, the divided moving image data 120 ′, 120 ″, 120 ′ ″ output from the frequency resolving unit 2 is appropriately changed according to the wavelet expansion coefficient ω _k ^(j) and the scaling coefficient s _k ^(j) . Is added and reconfigured.

For this reason, each of the frame images 130 after the inverse transformation by the image reconstruction unit 4 is time-dependent by deleting or reducing the wavelet expansion coefficient ω _k ^(j) depending on the time axis T by the coefficient changing unit 3. A noise component having a relatively low frequency is eliminated. Further, by enlarging the wavelet expansion coefficient ω _k ^(j) indicating a high-frequency component that does not depend on the time axis T, each frame image 130 after being inversely transformed by the image reconstruction unit 4 is a clear image with emphasized edges. It becomes.

  Next, experimental examples and results for realizing the above-described configuration and method will be described. FIG. 7 shows a schematic configuration of an ultrasonic echo image system including the image processing apparatus 100. In FIG. 7, S is an object to be used as a sample when acquiring an ultrasonic image. We use pork forelimbs to analyze fat, muscle and bone boundaries. The object S is immersed in the water W introduced into the water tank 201 which is a container. This is water that prevents the soft tissue of the object S from being deformed by the measuring force of the ultrasonic probe 205 described later. This is because the immersion method is applied. Reference numeral 202 denotes a water temperature meter that detects the temperature of the water W, and 203 denotes a heater that is placed in the water W. A heater control device (not shown) sets the temperature of the water W to, for example, 36 based on temperature information from the water temperature meter 202. The heating amount of the heater 203 is controlled so as to maintain the temperature.

  Reference numeral 205 denotes an ultrasonic probe provided toward the object S, and the probe 205 is attached to a chuck 207 of an NC machine tool used as an actuator via a probe holding device 206 corresponding to a fixing jig. It is done. As a result, the ultrasonic probe 205 can be moved toward or away from the object S with the rotation of the chuck 207. The reproducibility of the resulting image is taken into consideration.

  The ultrasonic probe 205 is connected to an ultrasonic imaging device 211 that generates an ultrasonic image of the object S via a cable 212. This device 211 is a low-priced machine developed mainly for the purpose of health care for measuring the thickness of the subcutaneous fat layer of a living body, and the maximum depth that can be measured is 105 mm as an example. Further, the ultrasonic probe 205 used here has 64 elements having a center frequency of 6 MHz, and has an effective measurement width of 37 mm.

  The ultrasonic imaging apparatus 211 is connected via a cable 213 to a computer (PC) 221 that embodies the hardware configuration of the image processing apparatus 100. As is well known, the PC 221 includes a control processing unit 223 including a CPU and a storage device, an operation unit 224 capable of key operation, an input interface 225 for capturing an ultrasonic image from the ultrasonic image device 211, an LED, and the like. The image input unit 1 corresponds to the input interface 225, the frequency resolution unit 2, the coefficient change unit 3, and the image reconstruction unit 4 correspond to the control processing unit 223, and the image output unit 5 includes the display unit 226. This corresponds to the display unit 226. Note that the hardware configuration of the image processing apparatus 100 is not limited to that shown in FIG.

  In the above configuration, in the experiment, the table speed of the NC machine tool and hence the feed speed of the ultrasonic probe 205 are set to 153.692 mm / min, and an ultrasonic image of the object S is acquired by the ultrasonic imaging device 211 at a sampling time of 130 mSec. did. FIG. 8 shows a result of capturing these ultrasonic images into the PC 221 and outputting a reconstructed frame image 130 from the original frame image 110 using the image processing apparatus 100.

  In the figure, reference numeral 120 denotes original moving image data taken into the control processing unit 223 of the PC 221. As described above, this is composed of a collection of frame images 110 acquired at each sampling time along the time axis T. FIG. 8A shows one frame image 110 before wavelet transform.

  FIG. 8B is an example in which the three-dimensional wavelet analysis is performed by the frequency resolving unit 2, and the horizontal direction in the drawing is the time axis T, and the vertical direction is the vertical axis Y in one frame image 110. The coefficient changing unit 3 deletes the component of the time-dependent divided moving image data 120A, doubles the component of the divided moving image data 120B indicating the time-independent high-frequency component, and reconstructs this. The data is sent to the unit 4, and the analysis image is inversely transformed.

  FIG. 8C shows a reconstructed frame image 130 obtained by such a procedure. Compared with the original frame image 110 shown in FIG. 8A, it can be seen that the tissue boundary of the object S is emphasized and that the noise disappears.

  FIG. 9 is an enlarged view of the main part 150 of the original frame image 110 shown in FIG. 8A and the main part 160 of the re-converted frame image 130 shown in FIG. 8C. It can be seen that the main parts 150 and 160 are at the same position, and the noise N present in the original frame image 110 is removed in the frame image 130 after the reconversion.

The coefficient changing unit 3 of this example performs a process of deleting all the expansion coefficients (that is, replacing the wavelet expansion coefficient ω _k ^(j) with 0 ⁾ for the time-dependent divided moving image data 120A. For example, in an echo image of a constantly moving organ such as the heart, it is expected that the same processing is not necessarily appropriate. In this case, not all the expansion coefficients of the time-dependent divided moving image data 120A are used. The expansion coefficient of the target specific frequency component may be deleted (erased) or reduced. Thereby, even if it is the target object S which always moves, the noise component of the specific frequency which exists in the some frame image 110 can be removed effectively.

  As described above, the image processing apparatus 100 according to the present embodiment inputs an image as an input unit that sequentially inputs the frame images 110 constituting the moving image data 120 as a moving image along the time, that is, the time axis T. Section 1, frequency decomposition section 2 as a decomposing means for dividing moving image data 120 into a plurality of frequency bands, and time axis T among divided moving image data 120 ′, 120 ″, 120 ′ ″. The coefficient changing unit 3 as changing means for deleting or reducing the components of the divided moving image data 120A depending on the frequency change unit 2, and the processing opposite to the frequency decomposing unit 2 is performed based on the components deleted or reduced by the coefficient changing unit 3. As a result, all the moving image data 120 ′, 120 ″, 120 ′ ″ divided by the frequency resolving unit 2 are reconstructed, and the reconstructed moving image data 120 is reconstituted every time. The image reconstruction unit 4 as a reconstruction unit that generates a restored frame image 130 and the image output unit 5 as an output unit that outputs the frame image 130 restored by the image reconstruction unit 4 are provided. ing.

  Here, each frame image 110 constituting the moving image data 120 is not decomposed into a plurality of frequency bands, but the three-dimensional moving image data 120 itself is decomposed into a plurality of frequency bands. Then, with respect to the divided moving image data 120A depending on the time axis T, an appropriate frequency component is arbitrarily selected, and the divided moving image data 120 ′, 120 ″, 120 is deleted or reduced. If '' 'is reconstructed, it becomes possible to effectively remove noise components of specific frequencies existing in the plurality of frame images 110.

  Further, the coefficient changing unit 3 according to the present embodiment splits the moving image data 120B indicating a high-frequency component that does not depend on the time axis T among the divided moving image data 120 ′, 120 ″, 120 ′ ″. If all the divided moving image data 120 ′, 120 ″, 120 ′ ″ are reconstructed, the restored individual frame image 130 is more clearly enhanced with the edge emphasized. Can be obtained.

  In addition, the frequency resolving unit 2 in the present embodiment performs a wavelet transform process for dividing a band into a low frequency and a high frequency for each direction of the horizontal axis X, the vertical axis Y, and the time axis T of the moving image data 120. The image reconstruction unit 4 is a frequency resolving unit for each of the horizontal axis X, the vertical axis Y, and the time axis T of all the divided moving image data 120 ′, 120 ″, 120 ′ ″. 2 has a configuration that performs inverse wavelet transform processing.

  By doing this, it is possible to use widely known wavelet transform and wavelet inverse transform processes.

The above-mentioned component is a wavelet expansion coefficient ω _k ^(j) obtained by performing wavelet transform processing, and a sharper frame image 130 is output simply by changing the wavelet expansion coefficient ω _k ^(j). It becomes possible to do.

  Furthermore, if the wavelet transform process by the frequency resolving unit 2 and the wavelet inverse transform process by the image reconstruction unit 4 are performed using the Daubecheies 10 mother wavelet as in this embodiment, the final frame image 130 is obtained. Therefore, excellent image reconstruction can be realized and good resolution can be obtained.

  In the present embodiment, not only the configuration of the image processing apparatus 100 but also the method of executing the process by the image processing apparatus 100 can obtain the same effects. A preferred image processing method includes a step performed by the image input unit 1, a step performed by the frequency resolving unit 2, a step performed by the coefficient changing unit 3, a step performed by the image reconstruction unit 4, and a step performed by the image output unit 5. By performing these steps in order, it is possible to effectively remove noise components of specific frequencies existing in the plurality of frame images 110.

  The present invention is also achieved by other embodiments. That is, a recording medium (or storage medium) that records a program code of software that realizes the functions of the above-described embodiments is supplied to an arbitrary system or apparatus. Then, a computer (CPU or MPU) incorporated in the system or apparatus reads and executes the program code stored in the recording medium. In this case, the program code itself read from the recording medium realizes the functions of the above-described embodiment, and the recording medium on which the program code is recorded constitutes the present invention.

  Further, by executing the program code read by the computer, an operating system or the like running on the computer performs part or all of the actual processing based on the instruction of the program code. Needless to say, the present invention also includes the case where the functions of the above-described embodiments are realized by the processing.

  Furthermore, it is assumed that the program code read from the recording medium is written in a memory provided in a function expansion code inserted into the computer or a function expansion unit connected to the computer. Thereafter, based on the instruction of the program code, the function expansion card, the CPU provided in the function expansion unit, or the like performs part or all of the actual processing. Needless to say, the present invention also includes the case where the functions of the above-described embodiments are realized by the processing.

  When the present invention is applied to the recording medium, program code corresponding to the method described above is stored in the recording medium.

  In addition, this invention is not limited to the said Example, A various deformation | transformation implementation is possible in the range of the summary of this invention. In the above-described embodiment, the proposal regarding the sharpening of the ultrasonic image has been made, but the present invention contributes to the sharpening of all other images. Therefore, the industrial application field to which the present invention pertains extends not only to the ultrasonic echo device manufacturing industry but also to the moving image analysis device manufacturing industry.

1 Image input unit (input means)
2 Frequency resolving part (decomposing means)
3 Coefficient change part (change means)
4 Image reconstruction unit (reconstruction means)
5 Image output unit (output means)
100 Image processing apparatus

Claims (8)

Input means for sequentially inputting each frame image constituting the moving image;
Decomposition means for dividing the moving image into a plurality of frequency bands;
Change means for erasing or reducing the expansion coefficient of the target frequency component among the divided moving image components depending on time among the divided moving images,
Based on the erased or reduced components, the reverse processing of the disassembling means is performed to reconstruct all the divided moving images, and the reconstructed moving images are divided by time and restored. Reconstructing means for generating a frame image,
An image processing apparatus comprising: output means for outputting the restored frame image.   The image processing apparatus according to claim 1, wherein the changing unit enlarges a divided moving image component indicating a high-frequency component independent of time in the divided moving image. The decomposing means performs a wavelet transform process for each of the vertical, horizontal, and time directions of the moving image,
The image processing apparatus according to claim 1, wherein the reconstruction unit performs wavelet inverse transformation processing in each of vertical, horizontal, and time directions of all the divided moving images.   The image processing apparatus according to claim 3, wherein the component is a wavelet expansion coefficient obtained by performing the wavelet transform process.   The image processing apparatus according to claim 3, wherein the wavelet transform process by the decomposing unit and the wavelet inverse transform process by the reconstruction unit are performed using a Daubecheies 10 mother wavelet. An input step of sequentially inputting each frame image constituting the moving image;
A decomposition step of dividing the moving image into a plurality of frequency bands;
Among the divided moving images, among the divided moving image components depending on time, a changing step of deleting or reducing the expansion coefficient of the target frequency component;
Based on the erased or reduced components, the reverse processing of the disassembling means is performed to reconstruct all the divided moving images, and the reconstructed moving images are divided by time and restored. A reconstruction process for generating a frame image,
And an output step of outputting the restored frame image.   A program causing a computer to execute the image processing method according to claim 6.   A computer-readable storage medium storing the program according to claim 7.