CN104156917A - X-ray CT image enhancement method based on double energy spectrums - Google Patents

Abstract

The invention relates to the technical field of CT and provides an X-ray CT image enhancement method based on double energy spectrums. In order to realize image fusion and enhancement of two images and compensate defects of a single image, the invention adopts the technical scheme as follows: the X-ray CT image enhancement method based on double energy spectrums comprises the following steps: scanning to obtain multiple sets of dark field images and calculating the average value of dark fields; moving a sample out of a view field, and respectively obtaining N sets of bright field images under high voltage and low voltage; calculating projected images under the high voltage and low voltage; unifying projection data under the high voltage and low voltage to a dimension according the gray level unifying principle; performing wavelet decomposition on N sets of corrected high and low energy projected images at the same position, and performing image fusion through the wavelet transform method; performing CT reconstruction through the utilization of fused projected images. The X-ray CT image enhancement method is mainly used for CT equipment design and manufacturing.

Description

Based on the X ray CT image enchancing method of dual intensity spectrum

Technical field

The present invention relates to CT technical field, particularly a kind of X ray CT image enchancing method based on dual intensity spectrum.。

Technical background

Because the detected object of X ray CT is very complicated, and in imaging process, be subject to the impact of the many factors such as ray energy, scattering, noise, picture quality there are differences and have decline in various degree.Especially when in sweep object, the strong and weak difference of different directions yardstick difference absorption excessive or material is excessive, CT image quality degradation, conventionally performance is to reconstruct parts of images, and parts of images is hidden in None-identified among background, there is the phenomenon of image deterioration and loss in detail in the CT image of rebuilding under single power spectrum.

It is a hot fields of CT imaging research that CT image is strengthened.But be confined to the enhancing for CT image under single energy more.

Summary of the invention

In order to overcome the deficiencies in the prior art, realize two width images and carry out image co-registration enhancing, make up the defect of single image.For this reason, the technical scheme that the present invention takes is, based on the X ray CT image enchancing method of dual intensity spectrum, to comprise the steps:

Many group darkfield images are obtained in scanning, and calculate the mean value of described details in a play not acted out on stage, but told through dialogues;

Sweep object is placed on monitor station, the condition of scanning and maximum scan angle number are set;

At same position place, under different voltage parameters, under high-low voltage, respectively gather one group of fluoroscopy images respectively;

Turntable is rotated to default angle, continue to repeat previous step process, until number of times reaches maximum scan number of times;

Sample is shifted out to visual field, obtain respectively the N group bright-field image under high-low voltage, and calculate the bright field mean value under described high-low voltage;

To the element marking that lost efficacy of the bright field data under high-low voltage;

Inactive pixels in bright-field image under the fluoroscopy images obtaining under high voltage and high voltage is carried out to mark, and utilize interpolation method to revise inactive pixels;

Inactive pixels in bright-field image under the fluoroscopy images obtaining under low-voltage and low-voltage is carried out to mark, and utilize interpolation method to revise inactive pixels;

Calculate the projected image under high-low voltage;

Based on the unified principle of gray scale, by data for projection unification to yardstick under high voltage, low-voltage;

Carry out wavelet decomposition to the high and low energy projected image after the same position place calibration of N group, and carry out image co-registration by the method for wavelet transformation;

Utilize the projected image after merging to carry out CT reconstruction.

One group of fluoroscopy images of each collection under different voltage parameters respectively, twice voltage parameter is respectively 100KV and 40 KV.

To the identification of break bounds pixel in details in a play not acted out on stage, but told through dialogues situation, determine bad pixel by mean value and the variance of darkfield image, if the variance of pixel corresponding to darkfield image and darkfield image average is outside 3 times of standard deviation scopes, be labeled as inactive pixels;

The identification of break bounds pixel in bright field situation: mean value and variance by bright-field image are determined bad pixel, for same location of pixels, the standard deviation of N scanning of calculating pixel, if i, i=1, ..., the bright-field image that N scanning is obtained outside 3 times of standard deviation scopes, is labeled as inactive pixels with the difference of average bright-field image;

With the identification of the strong pixel that changes output non_uniform response of stream, the N group bright field data under different electric currents are obtained in scanning, simulate each pixel with the strong slope changing of stream, calculate and evaluate slope and standard deviation, if the difference of the slope of certain pixel and standard deviation, outside 3 times of standard deviation scopes, is labeled as inactive pixels.

Based on the unified principle of gray scale, by data for projection unification to yardstick under high voltage, low-voltage, concrete steps are as follows:

Calculate scaling, scaling equals under high energy the ratio of all projection sums under all projection sums and low energy;

High energy data for projection is processed, itself and scaling are divided by, the tonal range of the projected image under high-low voltage is on a yardstick like this.

Carry out image co-registration to the high and low energy image after the same position place calibration of N group by the method for wavelet transformation, concrete steps comprise:

The two width projected images that obtain under same position place high-low voltage are carried out to M layer wavelet decomposition, obtain respectively its high-frequency sub-band and low frequency sub-band;

Determine high frequency fusion rule, take weighted method to merge to high fdrequency component.First the high frequency coefficient in the same direction of same one-level is carried out to piecemeal, choosing point block size is W × W, wherein W is the number of partitioned matrix element, calculates the average gradient of two width image high-frequency information corresponding blocks, determines the fusion weight of all directions high frequency coefficient according to the ratio of average gradient;

Determine low frequency fusion rule, take the weighted method that significantly value is relevant to merge to low frequency component, first determine significantly value S, make S=(h*C _m) ²+ (v*C _m) ²+ (d*C _m) ²wherein h, v, d is respectively high frequency level, the vertical and diagonal coefficient that M level is decomposed, C _mbe the low frequency coefficient that M level is decomposed, * represents convolution; After image under high-low voltage is all carried out to significantly value calculating, determine the fusion weight of low frequency coefficient according to the ratio of remarkable value;

The high-frequency sub-band of fused images;

The low frequency sub-band of fused images;

By the high-frequency sub-band having merged and low frequency sub-band, by wavelet inverse transformation, carry out Image Reconstruction, obtain the image after merging;

Utilize the method for information entropy to assess the image after merging, stop if satisfying condition; If do not satisfied condition, adjust decomposed class, merge weight, re-start, until satisfy condition.

Compared with the prior art, technical characterstic of the present invention and effect:

Due to the attenuation coefficient and energy correlation of material, therefore the present invention can obtain the distinctive attenuation change of material by the energy spectrum that changes x-ray source.Carry out image co-registration enhancing for two width images in dual intensity spectrum situation, can make up the defect of single image, significant.

Brief description of the drawings

Fig. 1 is the overall flow figure that an embodiment is carried out to figure image intensifying;

Fig. 2 is the process flow diagram that the image to obtaining under high and low energy of an embodiment merges.

Embodiment

The object of the present invention is to provide a kind of X ray CT image enchancing method based on dual intensity spectrum.The method comprises the steps:

Many group darkfield images are obtained in scanning, and calculate the mean value of described details in a play not acted out on stage, but told through dialogues;

Sweep object is placed on monitor station, the condition of scanning and maximum scan angle number are set;

At same position place, under different voltage parameters, under high-low voltage, respectively gather one group of fluoroscopy images respectively;

Turntable is rotated to default angle, continue to repeat previous step process, until number of times reaches maximum scan number of times;

Sample is shifted out to visual field, obtain respectively the N group bright-field image under high-low voltage, and calculate the bright field mean value under described high-low voltage;

To the element marking that lost efficacy of the bright field data under high-low voltage;

Inactive pixels in bright-field image under the fluoroscopy images obtaining under high voltage and high voltage is carried out to mark, and utilize interpolation method to revise inactive pixels;

Inactive pixels in bright-field image under the fluoroscopy images obtaining under low-voltage and low-voltage is carried out to mark, and utilize interpolation method to revise inactive pixels;

Calculate the projected image under high-low voltage;

Based on the unified principle of gray scale, by data for projection unification to yardstick under high voltage, low-voltage;

Carry out wavelet decomposition to the high and low energy projected image after the same position place calibration of N group, and carry out image co-registration by the method for wavelet transformation;

Utilize the projected image after merging to carry out CT reconstruction;

In one embodiment, one group of fluoroscopy images of each collection under different voltage parameters respectively, twice voltage parameter is respectively 100KV and 40KV;

In one embodiment, many group darkfield images are obtained in scanning, and calculate the mean value of described details in a play not acted out on stage, but told through dialogues the N group bright-field image under high and low energy is obtained in scanning, and calculates the mean value of described bright field, wherein for the evaluation of estimate of bright field in high energy situation, H represents high energy situation, for the evaluation of estimate of bright field in low energy situation, L represents low energy situation;

In one embodiment, to the identification of break bounds pixel in details in a play not acted out on stage, but told through dialogues situation, mean value and variance by darkfield image are determined bad pixel, if the variance of pixel corresponding to darkfield image and darkfield image average is outside 3 times of standard deviation scopes, are labeled as inactive pixels;

The identification of break bounds pixel in bright field situation.Mean value and variance by bright-field image are determined bad pixel, for same location of pixels, the standard deviation of N scanning of calculating pixel, if i is (i=1, ..., N) bright-field image that inferior scanning is obtained outside 3 times of standard deviation scopes, is labeled as inactive pixels with the difference of average bright-field image;

With the identification of the strong pixel that changes output non_uniform response of stream.The N group bright field data under different electric currents are obtained in scanning, simulate each pixel with the strong slope changing of stream, calculate and evaluate slope and standard deviation, if the interpolation of the slope of certain pixel and standard deviation is outside 3 times of standard deviation scopes, are labeled as inactive pixels;

In one embodiment, based on the unified principle of gray scale, by data for projection unification to yardstick under high energy, low energy, concrete steps are as follows:

Calculate scaling, scaling equals under high energy the ratio of all projection sums under all projection sums and low energy;

High energy data for projection is processed, itself and scaling are divided by, the tonal range of the projected image under high-low voltage is substantially on a yardstick like this.

In one embodiment, carry out image co-registration to the high and low energy image after the same position place calibration of N group by the method for wavelet transformation, concrete steps comprise:

The two width projected images that obtain under same position place high-low voltage are carried out to M layer wavelet decomposition, obtain respectively its high-frequency sub-band and low frequency sub-band;

Determine high frequency fusion rule, take weighted method to merge to high fdrequency component.First the high frequency coefficient in the same direction of same one-level is carried out to piecemeal, choosing point block size is W × W, calculates the average gradient of two width image high-frequency information corresponding blocks, determines the fusion weight of all directions high frequency coefficient according to the ratio of average gradient;

Determine low frequency fusion rule, take the weighted method that significantly value is relevant to merge to low frequency component.First determine significantly value S, make S=(h*C _m) ²+ (v*C _m) ²+ (d*C _m) ²wherein h, v, d is respectively high frequency level, the vertical and diagonal coefficient that M level is decomposed, C _mbe the low frequency coefficient that M level is decomposed, * represents convolution.After image under high-low voltage is all carried out to significantly value calculating, determine the fusion weight of low frequency coefficient according to the ratio of remarkable value;

The high-frequency sub-band of fused images;

The low frequency sub-band of fused images;

By the high-frequency sub-band having merged and low frequency sub-band, by wavelet inverse transformation, carry out Image Reconstruction, obtain the image after merging;

Utilize the method for information entropy to assess the image after merging, stop if satisfying condition; If do not satisfied condition, adjust decomposed class, merge weight, re-start, until satisfy condition.

For X ray CT image is strengthened, present embodiment provides a kind of X ray CT image enchancing method based on dual intensity spectrum, below in conjunction with example, the method is specifically described.

Please refer to Fig. 1, the concrete steps of present embodiment are as follows:

Close radiographic source, start detector, gather N width darkfield image, ask its average image to be designated as

Sweep object is placed on monitor station, the condition of scanning and maximum scan angle number N are set;

At same position place, each one group of fluoroscopy images, respectively note of gathering under different voltage parameters respectively the wherein counting of K mark diverse location, H represents high energy situation, L represents low energy situation;

Turntable is rotated to default angle, continue to repeat said process, until number of times reaches maximum scan times N;

Sample is shifted out to visual field, obtain respectively the many groups bright-field image under high energy, low energy, remember that its average image is respectively

Inactive pixels is carried out to mark, and the inactive pixels here comprises two parts: a part is break bounds pixel, the unstable pixel of another part.

To the identification of break bounds pixel in details in a play not acted out on stage, but told through dialogues situation, determine bad pixel by mean value and the variance of darkfield image, if the variance of pixel corresponding to darkfield image and darkfield image average is outside 3 times of standard deviation scopes, be labeled as inactive pixels;

Mean value and variance by bright-field image are determined bad pixel, for same location of pixels, the standard deviation of N scanning of calculating pixel, if i is (i=1, ..., N) bright-field image that inferior scanning is obtained outside 3 times of standard deviation scopes, is labeled as inactive pixels with the difference of average bright-field image;

With the strong identification that changes the corresponding inconsistent pixel of output of stream.The N group bright field data under different electric currents are obtained in scanning, simulate each pixel with the strong slope changing of stream, calculate and evaluate slope and standard deviation, if the interpolation of the slope of certain pixel and standard deviation is outside 3 times of standard deviation scopes, are labeled as inactive pixels;

According to bad element marking above, to the fluoroscopy images obtaining under high-low voltage bad pixel proofread and correct.Set up the bad picture element interpolation scheme based on bad template pixel, if bad pixel is single isolated point, carries out interpolation by its 3*3 neighborhood and fill up; If bad pixel is non-isolated point, and its around in 3*3 neighborhood bad number of pixels be greater than 4, adopt interpolation from outside to inside to fill up; If bad pixel is non-isolated point, and its around in 3*3 neighborhood bad number of pixels be less than 4, with around in 3*3 neighborhood good pixel fill up by distance weighted average interpolation.

Calculate projected image, $P_H^K(i,j)=\log \left(\frac{\stackrel{\‾}{I_H^O}(i,j)-\stackrel{\‾}{I_N}(i,j)}{I_H^K(i,j)-\stackrel{\‾}{I_N}(i,j)}\right),P_L^K(i,j)=\log \left(\frac{\stackrel{\‾}{I_L^0}(i,j)-\stackrel{\‾}{I_N}(i,j)}{I_L^K(i,j)-\stackrel{\‾}{I_N}(i,j)}\right),$ Wherein for the average image of N width darkfield image, be respectively the average of bright field under high energy, low energy, the fluoroscopy images gathering under high and low energy respectively.

Based on the unified principle of gray scale, the gray scale unification of the projected image under high low energy is arrived to same range scale;

Calculate scaling, be designated as Scale, $\mathrm{Scale}=\mathrm{\Σ}{P}_H^K(i,j)/\mathrm{\Σ}{P}_L^K(i,j);$

Order tonal range under high and low like this energy is substantially on a yardstick;

Carry out wavelet decomposition to the high and low energy projected image after the calibration of the N of same position place group, and carry out image co-registration by the method by wavelet transformation;

In one embodiment, carry out image co-registration to the high and low energy image after the calibration of the N of same position place group by the method for wavelet transformation, concrete steps comprise:

To the high and low two width projected images that can obtain down carry out M layer wavelet decomposition, obtain respectively its high-frequency sub-band and low frequency sub-band;

Determine high frequency fusion rule, take weighted method to merge to high fdrequency component.First, the high frequency coefficient in the same direction of same one-level is carried out to piecemeal, choosing point block size is W × W, calculates the average gradient of two width image high-frequency information corresponding blocks, determines the fusion weight of all directions high frequency coefficient according to the ratio of average gradient;

Determine low frequency fusion rule, take the weighted method that significantly value is relevant to merge to low frequency component.First determine significantly value S, make S=(h*C _m) ²+ (v*C _m) ²+ (d*C _m) ²wherein h, v, d is respectively high frequency level, the vertical and diagonal coefficient that M level is decomposed, C _mbe the low frequency coefficient that M level is decomposed, * represents convolution.After high and low energy image is all carried out to significantly value calculating, determine the fusion weight of low frequency coefficient according to the ratio of remarkable value;

The high-frequency sub-band of fused images;

The low frequency sub-band of fused images;

By the high-frequency sub-band having merged and low frequency sub-band, by wavelet inverse transformation, carry out Image Reconstruction, obtain the image after merging;

Utilize the method for information moisture in the soil to assess the area-of-interest of the image after merging, stop if satisfying condition; If do not satisfied condition, readjust the fusion coefficients of high frequency and low frequency, and decomposed class, re-start.

Adopt the image merging after strengthening to carry out image reconstruction.

Claims (5)

1. the X ray CT image enchancing method based on dual intensity spectrum, is characterized in that, comprises the steps:

Many group darkfield images are obtained in scanning, and calculate the mean value of described details in a play not acted out on stage, but told through dialogues;

Sweep object is placed on monitor station, the condition of scanning and maximum scan angle number are set;

At same position place, under different voltage parameters, under high-low voltage, respectively gather one group of fluoroscopy images respectively;

Turntable is rotated to default angle, continue to repeat previous step process, until number of times reaches maximum scan number of times;

Sample is shifted out to visual field, obtain respectively the N group bright-field image under high-low voltage, and calculate the bright field mean value under described high-low voltage;

To the element marking that lost efficacy of the bright field data under high-low voltage;

Inactive pixels in bright-field image under the fluoroscopy images obtaining under high voltage and high voltage is carried out to mark, and utilize interpolation method to revise inactive pixels;

Inactive pixels in bright-field image under the fluoroscopy images obtaining under low-voltage and low-voltage is carried out to mark, and utilize interpolation method to revise inactive pixels;

Calculate the projected image under high-low voltage;

Based on the unified principle of gray scale, by data for projection unification to yardstick under high voltage, low-voltage;

Carry out wavelet decomposition to the high and low energy projected image after the same position place calibration of N group, and carry out image co-registration by the method for wavelet transformation;

Utilize the projected image after merging to carry out CT reconstruction.

2. the X ray CT image enchancing method based on dual intensity spectrum as claimed in claim 1, is characterized in that, one group of fluoroscopy images of each collection under different voltage parameters respectively, and twice voltage parameter is respectively 100KV and 40 KV.

3. the X ray CT image enchancing method based on dual intensity spectrum as claimed in claim 1, it is characterized in that, to the identification of break bounds pixel in details in a play not acted out on stage, but told through dialogues situation, mean value and variance by darkfield image are determined bad pixel, if the variance of the pixel that darkfield image is corresponding and darkfield image average, outside 3 times of standard deviation scopes, is labeled as inactive pixels;

The identification of break bounds pixel in bright field situation: mean value and variance by bright-field image are determined bad pixel, for same location of pixels, the standard deviation of N scanning of calculating pixel, if i, i=1, ..., the bright-field image that N scanning is obtained outside 3 times of standard deviation scopes, is labeled as inactive pixels with the difference of average bright-field image;

With the identification of the strong pixel that changes output non_uniform response of stream, the N group bright field data under different electric currents are obtained in scanning, simulate each pixel with the strong slope changing of stream, calculate and evaluate slope and standard deviation, if the difference of the slope of certain pixel and standard deviation, outside 3 times of standard deviation scopes, is labeled as inactive pixels.

4. the X ray CT image enchancing method based on dual intensity spectrum as claimed in claim 1, is characterized in that, based on the unified principle of gray scale, by data for projection unification to yardstick under high voltage, low-voltage, concrete steps are as follows:

Calculate scaling, scaling equals under high energy the ratio of all projection sums under all projection sums and low energy;

High energy data for projection is processed, itself and scaling are divided by, the tonal range of the projected image under high-low voltage is on a yardstick like this.

5. the X ray CT image enchancing method based on dual intensity spectrum as claimed in claim 1, is characterized in that, carry out image co-registration to the high and low energy image after the same position place calibration of N group by the method for wavelet transformation, and concrete steps comprise:

The two width projected images that obtain under same position place high-low voltage are carried out to M layer wavelet decomposition, obtain respectively its high-frequency sub-band and low frequency sub-band;

Determine high frequency fusion rule, take weighted method to merge to high fdrequency component.First the high frequency coefficient in the same direction of same one-level is carried out to piecemeal, choosing point block size is W × W, wherein W is the number of partitioned matrix element, calculates the average gradient of two width image high-frequency information corresponding blocks, determines the fusion weight of all directions high frequency coefficient according to the ratio of average gradient;

Determine low frequency fusion rule, take the weighted method that significantly value is relevant to merge to low frequency component, first determine significantly value S, make S=(h*C _m) ²+ (v*C _m) ²+ (d*C _m) ²wherein h, v, d is respectively high frequency level, the vertical and diagonal coefficient that M level is decomposed, C _mbe the low frequency coefficient that M level is decomposed, * represents convolution; After image under high-low voltage is all carried out to significantly value calculating, determine the fusion weight of low frequency coefficient according to the ratio of remarkable value;

The high-frequency sub-band of fused images;

The low frequency sub-band of fused images;

By the high-frequency sub-band having merged and low frequency sub-band, by wavelet inverse transformation, carry out Image Reconstruction, obtain the image after merging;

Utilize the method for information entropy to assess the image after merging, stop if satisfying condition; If do not satisfied condition, adjust decomposed class, merge weight, re-start, until satisfy condition.