Dynamic range based ultrasound image correction system

Abstract

This invention relates to the field of image correction technology, specifically to an ultrasound image correction system based on dynamic range. The system includes a cepstral conversion module, used to extract ultrasound image scan line signals from an ultrasound probe, extract depth level sequences along the axial depth direction and convert them into logarithmic spectral level parameters to obtain a cepstral transform coefficient matrix, extract the amplitude and inverse frequency values ​​corresponding to the extreme points of equally spaced isolated peaks within the cepstral transform coefficient matrix to form feature vector parameters, and generate reverberation period peak fractions. In this invention, an adaptive truncation limiting variable is calculated based on the local dynamic range parameter set to extract the original coordinates of the center pixel grayscale level data, truncate and assign values ​​to establish a corrected output image matrix, and implement dynamic response precise truncation for speckle intensity in different regions. This effectively suppresses high-frequency speckle noise while maintaining the sharpness of edge contours, improving image contrast and spatial fidelity of the underlying data.

Description

Technical Field

[0001] This invention relates to the field of image correction technology, and more particularly to an ultrasound image correction system based on dynamic range. Background Technology

[0002] Image correction technology is dedicated to compensating for and reconstructing image quality degradation, distortion, and artifacts caused by physical limitations of equipment, external environmental interference, or internal system errors during image acquisition, transmission, or storage.

[0003] Current correction techniques in digital image processing rely on preset static global filtering thresholds or uniform spatial convolution kernels to process the entire pixel matrix. They employ uniform-scale masks to perform smoothing operations across highly non-uniform audio or optical signal fields, forcibly assimilating tissue boundaries with acoustic impedance differences. Static frequency transformation lacks adaptive spatial location determination when suppressing interference bands, resulting in indiscriminate attenuation of core feature bands. Uniform radiation intensity compensation logic is prone to causing signal oversaturation distortion in low signal-to-noise ratio regions, causing the repaired pixel matrix to lose the hierarchical transition features of its original physical dimensions. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and propose an ultrasound image correction system based on dynamic range.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A dynamic range-based ultrasound image correction system includes: The signal cepstral conversion module is used to extract the ultrasonic image scan line signal from the ultrasonic probe, extract the depth level sequence along the axial depth direction and convert it into logarithmic spectrum level parameters to obtain the cepstral transformation coefficient matrix, extract the amplitude and reciprocal frequency values ​​corresponding to the extreme points of equally spaced isolated peaks in the cepstral transformation coefficient matrix, form the feature vector parameters, and generate the reverberation period peak fraction. The artifact frequency band suppression module is used to extract the amplitude value based on the reverberation period peak fraction, compare the amplitude value with the threshold constant, extract the generated artifact frequency band position coordinates, and set the artifact frequency band position coordinates to zero by associating them with the cepstral domain frequency band coefficient value, thereby establishing an artifact-free spatial echo signal. The local speckle statistics module is used to set the sliding window size parameter according to the artifact-free spatial echo signal, extract the pixel gray level data within the sliding window, calculate the local speckle signal-to-noise ratio value, calculate the difference constant between the extreme values ​​corresponding to the pixel gray level data according to the sliding window size parameter, and obtain the local dynamic range parameter set. The image adaptive correction module is used to calculate an adaptive truncation limiting variable based on the local dynamic range parameter set, extract the original coordinates of the gray level data of the center pixel in the local dynamic range parameter set and the adaptive truncation limiting variable, truncate and assign values ​​to the extreme points at the original coordinates, and establish a corrected output image matrix.

[0006] Preferably, the step of obtaining the peak reverberation period fraction is as follows: The ultrasonic image scan line signal is extracted from the ultrasonic probe. The amplitude value of the ultrasonic image scan line signal is extracted point by point along the axial depth direction. The depth level sequence is constructed according to the order of the amplitude values. Logarithmic calculation is performed on the depth level sequence to convert the depth level sequence into logarithmic spectrum level parameters. Inverse Fourier transform operation is performed on the logarithmic spectrum level parameters in sequence to extract the real part data after transformation and obtain the cepstral transform coefficient matrix. Scan the cepstral transform coefficient matrix along the inverted frequency axis, calculate the difference slope between adjacent values ​​of the cepstral transform coefficient matrix, locate the coordinate point where the difference slope changes from positive to negative and mark it as an isolated peak extreme point, calculate the coordinate difference between adjacent isolated peak extreme points, select points with equal coordinate differences as equally spaced isolated peak extreme points, extract the absolute value of the amplitude and the inverted frequency axis coordinate value of the equally spaced isolated peak extreme points, and obtain the amplitude and inverted frequency value corresponding to the equally spaced isolated peak extreme points; The amplitude and reciprocal frequency values ​​corresponding to the isolated peak extreme points with equal intervals are arranged in ascending order of coordinates. The combined data column constitutes the feature vector parameters. The sum of squares of all values ​​in the feature vector parameters is calculated. The total number of values ​​contained in the feature vector parameters is counted. The sum of squares is divided by the total number of values ​​to obtain the quotient value. The quotient value is extracted as the peak score of the reverberation period.

[0007] Preferably, the step of obtaining the artifact frequency band position coordinates is as follows: The amplitude and reciprocal frequency values ​​within the peak fraction of the reverberation period are analyzed. When the amplitude value is greater than the threshold constant, the distribution trajectory of the reciprocal frequency value and the surface depth parameter of the acoustic probe are compared, the coordinates of the coincidence of the distribution trajectory are extracted, and the position coordinates of the artifact frequency band are generated.

[0008] Preferably, the steps for acquiring the artifact-free spatial echo signal are as follows: Based on the artifact frequency band position coordinates, read the cepstral domain frequency band coefficient value, find the index node corresponding to the artifact frequency band position coordinates within the cepstral domain frequency band coefficient value, modify the cepstral domain frequency band coefficient value associated with the index node to zero, and generate a zero-assignment relationship value. The zero-assignment correlation values ​​are spatially transformed and mapped to convert them into a time-dimensional sound wave sequence. The amplitude distribution of the sound wave sequence is extracted, the amplitude space matrix is ​​reconstructed, and an artifact-free spatial echo signal is established.

[0009] Preferably, the step of obtaining the local speckle signal-to-noise ratio value is as follows: Based on the artifact-free spatial echo signal, the length and width boundaries are defined to set the sliding window size parameters. The pixel grayscale level data within the sliding window are extracted by traversing the spatial coordinate system. The mean parameter and variance parameter are statistically analyzed and divided to obtain the local speckle signal-to-noise ratio value.

[0010] Preferably, the steps for obtaining the local dynamic range parameter set are as follows: Based on the local speckle signal-to-noise ratio value, the sliding window size parameter and pixel gray level data corresponding to the generation process are retrieved, the maximum and minimum extreme values ​​within the pixel gray level data are located, and the extreme value subtraction difference constant is calculated. Based on the extreme value subtraction difference constant, the corresponding dimension of local speckle signal-to-noise ratio is extracted by callback. The extreme value subtraction difference constant and the local speckle signal-to-noise ratio value are combined to form a set of mapping independent variable parameters. The data columns in the set are arranged to obtain the local dynamic range parameter set.

[0011] Preferably, the step of obtaining the adaptive truncation limiting variable is as follows: Traverse the data nodes of the local dynamic range parameter set, peel off the upper limit and lower limit of the data interval boundary corresponding to the mapping independent variable parameter set, extract the local speckle signal-to-noise ratio value associated with the local dynamic range parameter set, combine the upper limit of the data interval boundary, the lower limit of the data interval boundary and the local speckle signal-to-noise ratio value to generate the amplitude limiting evaluation base vector. Extract the upper limit of the interval boundary, the lower limit of the interval boundary, and the local speckle signal-to-noise ratio value within the baseline vector of the amplitude limiting evaluation. Determine the difference constant between extreme values ​​and calculate the adaptive truncation amplitude limiting variable.

[0012] Preferably, the step of obtaining the corrected output image matrix is ​​as follows: The original coordinates of the gray level data of the center pixel within the local dynamic range parameter set are located. The extreme points at the original coordinates are replaced and truncated according to the adaptive truncation and limiting variable. The weights are allocated according to the spatial distance ratio and the truncation level values ​​of adjacent histogram regions are cross-fused to establish a corrected output image matrix.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, the ultrasonic image scan line signal is extracted by an ultrasonic probe, and the depth level sequence is converted into logarithmic spectral level parameters along the axial depth direction to obtain the cepstral transform coefficient matrix. The amplitude and reciprocal frequency values ​​corresponding to the extreme points of equally spaced isolated peaks are extracted to form feature vector parameters, generating the reverberation period peak fraction. This quantifies the periodic reverberation characteristics in the sound wave transmission path, avoiding signal overlap interference in conventional time-domain analysis. Based on the reverberation period peak fraction, the amplitude values ​​and threshold constants are compared to extract the artifact frequency band position coordinates. The artifact frequency band position coordinates are associated with the cepstral domain frequency band coefficient values ​​and zeroed to establish an artifact-free spatial echo signal. This targets and eliminates periodic artifact energy at the core level of the frequency domain, preserving non-periodic real tissue echo information. Based on the artifact-free spatial echo signal, a sliding window is set to extract pixel grayscale level data. The difference constant between the local speckle signal-to-noise ratio and the extreme values ​​is calculated to obtain the local dynamic range parameter set, constructing a spatial statistical benchmark reflecting local tissue inhomogeneity. Based on the local dynamic range parameter set, the adaptive truncation and limiting variables are calculated. The original coordinates of the gray level data of the center pixel are extracted and truncated to establish a corrected output image matrix. Dynamic response and precise truncation are implemented for speckle intensity in different regions. While effectively suppressing high-frequency speckle noise, the sharpness of the edge contour is maintained, and the image contrast and spatial fidelity of the underlying data are improved. Attached Figure Description

[0014] Figure 1 This is a system flowchart of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0016] Please see Figure 1 The present invention provides a technical solution: an ultrasound image correction system based on dynamic range includes: The signal cepstral conversion module is used to extract the ultrasonic image scan line signal from the ultrasonic probe, extract the depth level sequence along the axial depth direction and convert it into logarithmic spectrum level parameters to obtain the cepstral transformation coefficient matrix, extract the amplitude and reciprocal frequency values ​​corresponding to the extreme points of equally spaced isolated peaks in the cepstral transformation coefficient matrix, form the feature vector parameters, and generate the reverberation period peak fraction. The artifact frequency band suppression module is used to extract the amplitude value based on the peak fraction of the reverberation period, compare the amplitude value with the threshold constant, extract the position coordinates of the generated artifact frequency band, and set the cepstral domain frequency band coefficient value associated with the artifact frequency band position coordinates to zero to establish an artifact-free spatial echo signal. The local speckle statistics module is used to set the sliding window size parameters based on the artifact-free spatial echo signal, extract the pixel gray level data within the sliding window, calculate the local speckle signal-to-noise ratio, calculate the difference constant between the extreme values ​​of the pixel gray level data based on the sliding window size parameters, and obtain the local dynamic range parameter set. The image adaptive correction module is used to calculate the adaptive truncation limiting variable based on the local dynamic range parameter set, extract the original coordinates of the gray level data of the center pixel within the adaptive truncation limiting variable and the local dynamic range parameter set, truncate and assign values ​​to the extreme points at the original coordinates, and establish the corrected output image matrix.

[0017] The steps to obtain the peak reverberation period score are as follows: The ultrasonic image scan line signal is extracted from the ultrasonic probe. The amplitude value of the ultrasonic image scan line signal is extracted point by point along the axial depth direction. The depth level sequence is constructed according to the order of the amplitude values. Logarithmic calculation is performed on the depth level sequence to convert the depth level sequence into logarithmic spectrum level parameters. Inverse Fourier transform operation is performed on the logarithmic spectrum level parameters in sequence to extract the real part data after transformation and obtain the cepstral transform coefficient matrix. Scan the cepstral transform coefficient matrix along the inverted frequency axis, calculate the difference slope between adjacent values ​​of the cepstral transform coefficient matrix, locate the coordinate point where the difference slope changes from positive to negative and mark it as an isolated peak extreme point, calculate the coordinate difference between adjacent isolated peak extreme points, select points with equal coordinate differences as equally spaced isolated peak extreme points, extract the absolute value of the amplitude and the coordinate value of the inverted frequency axis of the equally spaced isolated peak extreme points, and obtain the amplitude and inverted frequency values ​​corresponding to the equally spaced isolated peak extreme points. The amplitude and reciprocal frequency values ​​corresponding to the extreme points of isolated peaks are arranged in ascending order of coordinates. The combined data column constitutes the characteristic vector parameters. The sum of squares of all values ​​in the characteristic vector parameters is calculated. The total number of values ​​contained in the characteristic vector parameters is counted. The sum of squares is divided by the total number of values ​​to obtain the quotient value. The quotient value is extracted as the peak fraction of the reverberation period.

[0018] Specifically, based on the ultrasound image scan line signal extracted by the ultrasound probe, a sampling frequency of 50,000,000 times per second is set to continuously sample the echo waveform, converting the analog echo into discrete digital amplitude data. The amplitude values ​​of the ultrasound image scan line signal are extracted point-by-point along the axial depth direction. The depth sampling start point is set to the 0 point on the probe surface, and the sampling step size is set to 0.1 mm. The voltage envelope detection peak value corresponding to each step position is taken as the amplitude value of the current depth. A depth level sequence is constructed according to the order of the amplitude values, and the length of this sequence is limited to 1024 sampling points. If the actual number of sampling points is insufficient, zeros are added to the end of the sequence. Logarithmic calculation is performed on the depth level sequence, traversing all amplitude values ​​of the depth level sequence. Each amplitude value is added to a preset small bias constant, which is set to 0.000001. This value is calculated based on the minimum quantization noise floor parameter of the hardware analog-to-digital converter. When the amplitude value at the depth point is 0, the constant is added to eliminate the infinite abnormal data in the logarithmic operation. For each amplitude value after addition, the logarithm with the natural constant as the base is calculated, and the depth level sequence is converted into a logarithmic spectrum level parameter, generating a 1D array containing 1024 logarithmic values. The inverse Fourier transform operation is performed on the logarithmic spectrum level parameter in turn. The logarithmic values ​​in the 1D array are multiplied one by one by a Hanning window function of length 1024. The frequency truncation leakage edge at both ends of the signal is eliminated by multiplying point by point. The inverse fast Fourier transform algorithm is applied to process the windowed array, and a complex result sequence containing real and imaginary parts is calculated. The transformed real part data is extracted, and each complex element of the complex result sequence is traversed and only the real part is retained. They are rearranged according to the original index order. The above process is repeated line by line for all ultrasound scan lines. The 1D real number sequence generated by each scan line is sequentially concatenated into a 2D grid to obtain the cepstral transform coefficient matrix.

[0019] The cepstral transform coefficient matrix is ​​scanned along the inverted frequency axis. Data elements within the matrix are read column-by-column from top to bottom. The differential slope of adjacent values ​​in the cepstral transform coefficient matrix is ​​calculated. The amplitude of the previous inverted frequency position is subtracted from the amplitude of the next inverted frequency position and divided by the inverted frequency coordinate step size to locate the coordinate point where the differential slope changes from positive to negative. When the differential slope of the current position is determined to be greater than 0 and the differential slope of the next adjacent position is less than 0, the coordinate point is marked as an isolated peak extreme point. Before marking, a prominence check is introduced. The amplitude of the current coordinate point is extracted and subtracted from the lowest amplitude of the 10 adjacent data points on both sides to generate a prominence value. The prominence value is compared with a preset prominence threshold, which is set to 0.5. This setting value is calculated with reference to the energy level difference between the scattered echo of normal tissue and reverberation interference. For example, the prominence value generated is 0.8. If the difference is greater than 0.5, it is marked as an affirmation. The coordinate difference between adjacent isolated peak extreme points is calculated. The isolated peak extreme points marked as affirmation are sorted in ascending order of reciprocal frequency coordinates. The distance value is generated by subtracting the coordinate of the previous extreme point from the coordinate of the next extreme point. Points with equal coordinate differences are selected as equally spaced isolated peak extreme points. The distance tolerance range is set to ±3 sampling points. The first distance value is used as the reference value. If the subsequent distance falls within the reference value plus or minus 3 sampling points, for example, if the reference distance is 50, the subsequent distance is judged to be equal in coordinate difference when it is between 47 and 53. These consecutive extreme points are retained. The absolute value of the amplitude and the reciprocal frequency coordinate value of the equally spaced isolated peak extreme points are extracted. The negative sign of the retained extreme point amplitude value is removed and converted into an absolute value. The absolute value is then paired and integrated with the reciprocal frequency coordinate to obtain the amplitude and reciprocal frequency value corresponding to the equally spaced isolated peak extreme points.

[0020] Arrange the amplitude and reciprocal frequency values ​​corresponding to the extreme points of isolated peaks at equal intervals in ascending order of coordinates. Use a quicksort method to compare the magnitudes of all reciprocal frequency axis coordinate values, placing the smallest coordinate at the beginning and moving the remaining coordinates in ascending order. When moving coordinate positions, simultaneously move the bound absolute amplitude values ​​to maintain the original binding mapping relationship between coordinates and amplitudes. The combined and arranged data columns constitute the feature vector parameters. Create a blank 1D floating-point array, fill the odd-indexed positions of the array with the sorted reciprocal frequency coordinate values, and fill the even-indexed positions with the corresponding absolute amplitude values, concatenating them to form the 1D feature vector parameters. Traverse the vector to check for null values. If a missing data is found, replace it with the adjacent valid value. Calculate the sum of squares of all values ​​in the feature vector parameters, and set an initial value of 0 for the cumulative variable. The algorithm reads each data element in the feature vector parameters one by one, multiplies the read data element by itself to generate a square value, adds this square value to the current sum stored in the accumulation variable, and writes the new sum back into the accumulation variable. This reading, multiplication, and accumulation operation is repeated until the last element of the feature vector parameters is reached. After the traversal is completed, the value stored in the accumulation variable is the final sum of squares. The algorithm counts the total number of values ​​contained in the feature vector parameters, reads the total number of bytes occupied by the 1D array, divides it by the 4-byte length of a single floating-point data standard to calculate the total number of specific elements contained in the array, divides the previously calculated sum of squares by the total number of values ​​to obtain the quotient value, uses the sum of squares output by the accumulation variable as the dividend, and uses the total number of values ​​as the divisor to perform a division operation, extracting the quotient value as the peak score of the reverberation period.

[0021] The steps to obtain the location coordinates of the artifact frequency band are as follows: The amplitude and reciprocal frequency values ​​within the peak fraction of the reverberation period are analyzed. When the amplitude value is greater than the threshold constant, the distribution trajectory of the reciprocal frequency value and the surface depth parameter of the acoustic probe are compared, the coordinates of the coincidence of the distribution trajectory are extracted, and the position coordinates of the artifact frequency band are generated.

[0022] Specifically, the amplitude and reciprocal frequency values ​​within the peak fraction of the reverberation period are analyzed. The peak fraction data series is read, and the amplitude and reciprocal frequency values ​​associated with each data point are separated. A threshold constant is set as the judgment criterion. This threshold constant is specifically set by multiplying the average amplitude of the background noise obtained during the unobstructed water tank calibration experiment by a proportional coefficient. For example, in the pure water tank calibration test, 1000 consecutive samples were taken, and the average amplitude of the background noise was 0.15. With a proportional coefficient of 3.5, the calculated threshold constant is 0.525. Each amplitude value within the peak fraction of the reverberation period is extracted and compared with 0.525. If the currently read amplitude value is greater than 0.525, a strong reflection characteristic is determined at that reciprocal frequency value. The hardware factory configuration information of the ultrasonic detection equipment is retrieved, and the surface depth parameter of the sound wave probe is read. This depth parameter is established by combining the probe's physical dimensions with the sound wave propagation speed of 1540 meters per second in human soft tissue. The distribution trajectory array changes over time. Simultaneously, the inverse frequency values ​​corresponding to amplitude values ​​greater than a threshold constant are extracted. These inverse frequency values ​​are mapped to the same time scale to form inverse frequency distribution trajectories. The mapping conversion method involves multiplying the inverse frequency value by half the speed of sound and then dividing by the sampling frequency to obtain the equivalent spatial depth. This establishes the spatial depth distribution trajectory of the inverse frequency characteristics. The distribution trajectories of the inverse frequency values ​​and the surface depth parameters of the acoustic probe are compared. The time scale indices of these two trajectory arrays are traversed, and the absolute value of the difference between the inverse frequency trajectory value and the depth trajectory value at the same index is calculated one by one. A coincidence tolerance constant of 0.05 mm is set. When the calculated absolute value of the difference is less than or equal to 0.05 mm, it is determined that the two trajectories spatially overlap at the current coordinate point. The inverse frequency axis index position and the spatial depth coordinate axis index position at the time of overlap are recorded. These coordinate points that meet the coincidence condition are summarized and packaged, and the coincidence coordinates of the distribution trajectories are extracted to generate artifact frequency band position coordinates.

[0023] The steps for obtaining artifact-free spatial echo signals are as follows: Based on the artifact frequency band location coordinates, read the cepstral domain frequency band coefficient values, find the index node corresponding to the artifact frequency band location coordinates within the cepstral domain frequency band coefficient values, modify the cepstral domain frequency band coefficient values ​​associated with the index nodes to zero, and generate zero-assignment relationship values. Spatial transformation mapping is performed on the zero-assignment correlation values ​​to convert them into time-dimensional sound wave sequences. The amplitude distribution of the sound wave sequences is extracted, the amplitude space matrix is ​​reconstructed, and an artifact-free spatial echo signal is established.

[0024] Specifically, based on the artifact frequency band location coordinates, the complete cepstral transform coefficient data set pre-stored in memory is called to read the cepstral domain frequency band coefficient values. These cepstral domain frequency band coefficient values ​​are presented as a one-dimensional array composed of multiple discrete data points. The artifact frequency band location coordinate set is expanded, and the information of each coordinate point in the coordinate set is traversed to find the index node corresponding to the artifact frequency band location coordinate within the cepstral domain frequency band coefficient value. Since the artifact frequency band location coordinate contains relative position information of spatial depth, it is divided by the single depth sampling step size parameter. For example, if the single depth sampling step size is set to 0.1 mm, and the spatial depth indicated by the current artifact frequency band location coordinate is 5.5 mm, the quotient obtained by dividing the two is 55. This quotient is rounded down as the array subscript to locate the 55th index node in the one-dimensional array of cepstral domain frequency band coefficient values. To account for possible data offsets in coordinate mapping, the index lookup window range is set to ±2 nodes of the location node. Within the specified range, index nodes 53 to 57 are simultaneously locked as the target node interval to be processed. This covers the coordinate drift error caused by sampling jitter, completes the traversal and index calculation of all artifact frequency band position coordinates, and establishes a list of index nodes to be processed. This list is temporarily stored in the cache using a hash table data structure, with the key value being the array subscript. All nodes in the list are locked, and the cepstral array is accessed cyclically according to the key value information provided by the hash table. The cepstral domain frequency band coefficient values ​​associated with the index nodes are modified to zero. The frequency band coefficient values ​​within the locked range are overwritten using the array subscript pointer. Regardless of the magnitude of the original coefficient value, it is forcibly replaced with 0.0 to eliminate the periodic reverberation energy components caused by multiple reflections in this frequency band. The original cepstral domain frequency band coefficient values ​​of the index nodes outside the list that are not locked remain unchanged. The entire 1D array that has been overwritten and modified is re-encapsulated to generate zero-assignment related values.

[0025] The zero-assignment correlation values ​​are spatially transformed and mapped, transforming the cepstral sequence containing the zero-assignment data back to the conventional time or frequency domain. Homomorphic inverse calculations of the fast Fourier transform are performed on the 1D array of zero-assignment correlation values ​​to restore the processed cepstral data to a logarithmic spectrum. Then, an exponential function operation is used to unconvolve the logarithmic spectrum, removing the logarithmic compression effect and mapping the spectral data back to the original linear amplitude space. An inverse fast Fourier transform is performed to convert the frequency domain data back to the time domain, converting the zero-assignment correlation values ​​into a time-dimensional acoustic sequence. The output acoustic sequence is now a set of discrete data points of voltage amplitude varying over time. The amplitude distribution of the acoustic sequence is extracted, and the absolute amplitude of each discrete data point is read sequentially. The continuous time sequence is divided according to a set acoustic scan line pulse repetition frequency. For example, if the ultrasonic probe's pulse repetition frequency is set to 5000 times per second, each transmission forms a sequence containing 10... The time series of 24 sampling points is used to extract 1024 amplitude data points corresponding to each scan line and arrange them into independent column vectors. The total number of scan lines in the current frame image is retrieved. For example, an ultrasound image consists of 128 scan lines. These 128 column vectors are stitched together from left to right according to the physical excitation order of the probe array to reconstruct the amplitude space matrix, forming a two-dimensional matrix structure of 1024 rows by 128 columns. The row number in the matrix represents the distance coordinate along the axial depth direction, and the column number represents the width coordinate along the transverse horizontal direction of the probe. Each value filled in the matrix represents the echo reflection intensity at the corresponding spatial coordinate point after removing reverberation interference. A bilinear interpolation algorithm is introduced to perform spatial coordinate axis grid alignment and pixel gap interpolation smoothing on the reconstructed two-dimensional matrix data. The distance-weighted average of the values ​​of four adjacent known grid points is calculated to fill the blank areas. The smoothed complete two-dimensional matrix is ​​output to establish an artifact-free spatial echo signal.

[0026] The steps for obtaining the signal-to-noise ratio of local speckle are as follows: Based on the artifact-free spatial echo signal, the length and width boundaries are defined to set the sliding window size parameters. The pixel grayscale level data within the sliding window are extracted by traversing the spatial coordinate system. The mean and variance parameters are statistically analyzed and divided to obtain the local speckle signal-to-noise ratio value.

[0027] Specifically, based on the artifact-free spatial echo signal, the coordinates of all pixels and their corresponding amplitude values ​​within the two-dimensional amplitude space matrix are read. The length and width boundaries are defined to set the sliding window size parameters. The axial and lateral resolution parameters from the probe hardware configuration information are called. The sliding window size parameters are set with reference to the ratio between the ultrasonic length and the physical size of the speckle. In a standard abdominal ultrasound detection scenario, the physical width and height of the speckle patch are measured to be 1 mm and 1 mm, respectively. Combining this with the physical mapping ratio of image pixels (e.g., each pixel represents 0.2 mm), the number of pixels required to cover a complete speckle patch is calculated. Dividing 1 mm by 0.2 mm yields a width base of 5 pixels, and similarly, a height base of 5 pixels. The sliding window size is set to a rectangle with a width and height of 5 pixels. The spatial coordinate system is traversed to extract the pixel grayscale data within the sliding window. The set 5x5 rectangular sliding window is placed at the starting coordinate position of the upper left corner of the two-dimensional amplitude space matrix. The sliding window is moved pixel by pixel in a left-to-right and top-to-bottom order, ensuring that the window covers the entire image during the movement. The system operates within a spatial coordinate system and does not exceed the image's width and height boundaries. At each sliding pause, it reads the absolute amplitude values ​​of the 25 pixels contained within the current window, using these as the pixel grayscale level data for that local area. It then calculates the mean and variance parameters and performs a division operation, summing the grayscale data of these 25 pixels. This sum is divided by the number of pixels (25) to calculate the mean parameter. For example, if the sum of the 25 extracted pixel values ​​is 3000, the mean parameter calculated after division is 120. Finally, it subtracts this mean from the value of each pixel within the window. The mean parameter is set to 120 and squared. The sum of these 25 squared values ​​is then divided by 25 to calculate the variance parameter. The variance parameter is set to 400. The mean parameter of 120 is used as the dividend and the variance parameter of 400 is used as the divisor to perform a division operation. The division produces a ratio of 0.3. This extraction and statistical division calculation is performed iteratively for each pixel coordinate in the image matrix. All the calculated ratios are rearranged according to the original spatial coordinate mapping position to form a new matrix with the same size as the original image, thus obtaining the local speckle signal-to-noise ratio value.

[0028] The steps for obtaining the local dynamic range parameter set are as follows: Based on the local speckle signal-to-noise ratio, the sliding window size parameters and pixel grayscale level data corresponding to the generation process are retrieved. The maximum and minimum extreme values ​​within the pixel grayscale level data are located, and the difference constant between the extreme values ​​is calculated using the following formula: ; in, The difference constant between extreme values ​​is the constant value. This represents the maximum value of a pixel's grayscale level data. This represents the minimum extreme value within the pixel grayscale level data. This is the width base for the sliding window size parameter. The height base for the sliding window size parameters; Based on the difference constant between extreme values, the corresponding local speckle signal-to-noise ratio value is extracted via callback. The difference constant between extreme values ​​and the local speckle signal-to-noise ratio value are combined to form a set of mapping independent variable parameters. The data columns within the set are arranged to obtain the local dynamic range parameter set.

[0029] Specifically, in the formula for calculating the difference constant between extreme values, the exponential decay characteristic of the window size factor and the ratio of extreme values ​​is introduced to nonlinearly adjust the basic range. By using the geometric ratio of width and height and the weight of noise floor and peak value, the excessive correction amplitude of high contrast area under large window size is suppressed, and the true texture distribution of detail edges is preserved. The steps for obtaining the parameter are as follows: This parameter represents the maximum value of the pixel grayscale level data. To obtain this parameter, it is necessary to retrieve the grayscale level data of all pixels within the coverage area of ​​the current sliding window. Taking a 5x5 sliding window as an example, read the grayscale values ​​corresponding to the 25 pixels contained within it. Set the initial comparison variable to 0, iterate through these 25 values ​​and compare them one by one with the comparison variable. If the current value read is greater than the comparison variable, then the current value is written into the comparison variable. After completing 25 comparisons, extract the highest grayscale peak value. For example, after comparison and filtering, the highest grayscale value extracted is 240. Perform maximum and minimum normalization calculation on the extracted absolute value. Read the upper limit of grayscale allowed by the probe hardware, 255, and the lower limit of grayscale, 0. Apply the normalization calculation formula as follows: Substitute the extracted values ​​into the data for calculation. Dividing by this yields a dimensionless normalized value of 0.941. The specific value assigned is 0.941; The parameter acquisition steps are as follows: This parameter represents the minimum extreme value within the pixel grayscale level data. Similarly, read the grayscale level data of 25 pixels within the current 5x5 sliding window, setting the initial comparison variable to 255. Read each of these 25 grayscale values ​​and compare them with the comparison variable. If the currently read value is less than the comparison variable, replace and update the comparison variable with the smaller value. After completing 25 iterations of comparison operations, lock the darkest signal point of the local noise level. For example, select and extract the lowest value as 30. Apply maximum-minimum normalization calculation processing, using the same extreme value boundary parameters for transformation. The normalization calculation formula is set as follows: Substitute the found minimum value into the data to calculate. Dividing by this yields a dimensionless normalized noise floor value of 0.118. The specific value assigned is 0.118; The parameter acquisition steps are as follows: This parameter represents the width base number corresponding to the sliding window size parameter. Its value comes from the previously set horizontal pixel span of the window. It is calculated by reading the pre-configured probe horizontal resolution and speckle physical width in memory, combined with the horizontal coordinate interval of the digital image sampling mapping. In the aforementioned data extraction step, the number of horizontal pixels required to cover 1 mm of physical speckle has been defined. By dividing 1 mm by the 0.2 mm physical size represented by a single pixel, it is calculated that 5 pixels need to be covered horizontally, thus obtaining the width base number corresponding to the sliding window size parameter. The specific value assigned is 5; The parameter acquisition steps are as follows: This parameter represents the height base corresponding to the sliding window size parameter. This parameter is based on the axial detection depth resolution of the ultrasonic probe and the physical height setting of the speckle. It is calculated by reading the vertical axis sampling ratio written into the header file during the initial image construction. Using the standard requirement of covering 1 mm of physical speckle depth, 1 mm is divided by the 0.2 mm physical depth distance represented by a single axial pixel. The calculation results in a vertical dimension that must contain 5 consecutive pixels. This value defines the coverage length of the sliding window in the row direction of the image matrix. This height pixel number value is extracted as a spatial weight reference variable participating in the geometric ratio calculation of the formula to obtain the height base corresponding to the sliding window size parameter. The specific value assigned is 5; Calculations based on parameters: Substitute the aforementioned normalized and pixel count parameter values ​​into the formula. =0.941, =0.118, =5, =5; Calculate the window size scaling factor: ; Find the sum of squares: ; Find the product of the denominators: ; Calculate the result of the division and take the square root: ; Calculate the exponential decay factor: ; Calculate the power of the exponent: ; Calculate the natural constant raised to the power of 0.125: ; Calculate the baseline range value: ; Combining the products of each term yields the final result: ; The results indicate that the normalized extreme value subtraction difference constant, after nonlinear weighting by the spatial window ratio and extreme value attenuation within the current local image patch, reflects that the current region not only has high inherent contrast but also relatively low noise interference on the signal peak. The numerical result of 0.932 is used as a dimensionless mapping coefficient for subsequent scaling of the local speckle signal-to-noise ratio. If the value is close to 1.0, it indicates that the local area has strong scattering tissue interface features, requiring a larger truncation tolerance to prevent useful edges from being removed. If the value is below 0.4, it is mostly pure blood flow or cystic dark areas, guiding subsequent steps to shrink the dynamic range to suppress low-level random noise.

[0030] Based on the extreme value subtraction difference constant, the image coordinate record table is called to locate the pixel coordinates of the center point of the sliding window where the current calculation data is located. Based on the information of this spatial coordinate point, the corresponding dimension's local speckle signal-to-noise ratio (SNR) value is extracted via callback. For example, at the position of the 10th column horizontally and the 10th row vertically, the previously calculated local speckle SNR value is 0.3, and the corresponding dimensionless extreme value subtraction difference constant is 0.932. These two different dimensions of data extracted under the same spatial coordinates are packaged and bound together to establish a multi-dimensional data dictionary indexed by coordinate points. The extreme value subtraction difference constant and the local speckle SNR value are combined to form a set of mapping independent variable parameters. This binding and extraction operation is repeatedly performed on tens of thousands of sliding window center points in the entire two-dimensional image space. All parameter pairs are summarized into a two-dimensional data table in memory, where the first column stores the local speckle SNR value, and the second column stores the extreme value subtraction difference constant. The data columns within the set are arranged, and the two-dimensional data table is reconstructed using a bubble sort algorithm. The local speckle SNR value in the first column is set as the sorting criterion key, and sorted according to value from... All data rows are moved and swapped vertically in ascending order, maintaining the row binding mapping relationship between the first and second columns during the swapping process. Simultaneously, a reasonable range validation and filtering is performed on each row of data after sorting. Each record is compared to a pre-defined valid range; for example, the local speckle signal-to-noise ratio is compared to the valid range of 0.1 to 3.0, and the extreme value subtraction difference constant is compared to the reasonable mapping range of 0.0 to 1.5. These two ranges are established based on the historical echo statistics of the probe in human tissue. If the signal-to-noise ratio of a row of data is 0.05, lower than the lower limit of 0.1, or the extreme value subtraction difference constant is 1.8, exceeding the upper limit of 1.5, then the row of data is determined to be affected by transient interference from probe bubble shedding. The entire row containing this abnormal data is completely removed from the data table and destroyed to avoid abnormal extreme values ​​causing offset interference to the subsequent evaluation of truncated variables. After the complete sorting, cleaning, and filtering operations, the remaining valid and ordered two-dimensional data table is output and solidified into a structure format that can be read by subsequent steps to obtain the local dynamic range parameter set.

[0031] The steps for obtaining the adaptive truncation limit variable are as follows: Traverse the data nodes of the local dynamic range parameter set, peel off the upper and lower limits of the data interval boundary corresponding to the mapping independent variable parameter set, extract the local speckle signal-to-noise ratio value associated with the local dynamic range parameter set, combine the upper and lower limits of the data interval boundary and the local speckle signal-to-noise ratio value to generate the basic vector for amplitude limiting evaluation. Extract the upper and lower limits of the interval boundaries and the local speckle signal-to-noise ratio values ​​within the baseline vector for amplitude limiting evaluation. Obtain the difference constant between extreme values ​​and calculate the adaptive truncation amplitude limiting variable. The calculation formula is as follows: ; in, To adaptively truncate the amplitude limiting variable, To limit the upper bound of the interval boundary within the basic vector for evaluation, To limit the lower bound of the interval boundary within the baseline vector for amplitude assessment, To limit the amplitude and evaluate the local speckle signal-to-noise ratio within the base vector, The difference constant between extreme values ​​is the constant value. This is a preset reference grayscale level constant.

[0032] Specifically, the process iterates through the data nodes of the local dynamic range parameter set, reads the two-dimensional data table in memory that has been filtered for validity, and visits each data node record in the table row by row. It identifies the set of mapping independent variables attached to the current node, and for each node's pixel grayscale characteristics, it strips the upper and lower limits of the data interval boundaries corresponding to the mapping independent variable parameter set. It then retrieves the quantization bit width configuration parameters of the ultrasound equipment's analog-to-digital converter. For example, if the current scanner is configured with eight-bit unsigned integer data for the digital representation of ultrasound images, the physical lower limit of the absolute grayscale boundary is set to 0 and the physical upper limit to 255. To completely eliminate dimensional conflicts in subsequent addition operations and maintain the requirement of completely dimensionless processing, these two physical boundary values ​​are subjected to max-min normalization and unit stripping. Dividing 0 by 255 calculates the lower limit of the data interval boundary as 0.0, and dividing 255 by 255 calculates the upper limit of the data interval boundary as 1.0. After boundary stripping, according to the memory address pointer of the currently accessed node, it extracts the local speckle signal-to-noise ratio value associated with the local dynamic range parameter set. This process has already been completed in the aforementioned calculation steps. The specific scaling factor was calculated from the dimensionless pixel features. For example, the local speckle signal-to-noise ratio (SNR) at the node in the 10th column horizontally and the 10th row vertically was 1.2. This value was retrieved from the first column of the two-dimensional data table. A one-dimensional floating-point continuous array space was allocated in the memory cache. The upper limit of the data interval boundary, the lower limit of the data interval boundary, and the local speckle SNR value were combined. The normalized upper limit of the data interval boundary, 1.0, was placed at the first index position of the array, the normalized lower limit of the data interval boundary, 0.0, was placed at the second index position, and finally the local speckle SNR value, 1.2, was filled into the third index position. These three discrete and independent pure scaling values ​​were bundled and packaged into a three-dimensional array structure in this fixed order to generate the limiting evaluation basis vector. The constructed vector was temporarily stored in a high-speed register for the next stage of adaptive truncation calculation. The remaining data nodes were traversed in a loop to generate the corresponding evaluation basis vector for each row of the two-dimensional data table, thereby ensuring that each local region of the entire image has an independent boundary adjustment basis.

[0033] In the adaptive truncation limiting variable calculation formula, the square term of the dimensionless signal-to-noise ratio and the range constant is used as the dynamic interpolation weight. Nonlinear interpolation adjustment is performed between the upper and lower limits of the interval. When the signal-to-noise ratio and contrast of a local area are high, the evaluation of the limiting variable is biased towards the upper limit value, and vice versa. By introducing the reference gray level ratio constant as a smoothing damping factor, the calculation overflow or divergence is prevented in the low signal-to-noise ratio area due to the denominator approaching zero, so as to achieve adaptive truncation smoothing control for different acoustic scattering characteristics. The steps for obtaining the parameter are as follows: This parameter represents the upper limit of the interval boundary within the baseline vector for amplitude limiting evaluation. Obtaining this parameter requires retrieving the one-dimensional array structure constructed in the previous steps. Accessing the first index position of the array according to a fixed memory address offset, the value stored therein is read. This value comes from the normalized transformation result of the highest physical threshold of the pixel quantization range of the ultrasound image. In eight-bit image depth, the physical absolute upper limit is 255. Dividing itself by the maximum full-scale range of 255 and stripping the grayscale physical units, it is forcibly mapped to a dimensionless probability interval of 0 to 1. In this way, a pure upper limit weight scaling factor is calculated. This calculated value is extracted by reading the baseline vector for amplitude limiting evaluation, thus obtaining the parameter. The specific value assigned is 1.0; The parameter acquisition steps are as follows: This parameter represents the lower limit of the interval boundary within the clipping evaluation base vector. Its acquisition process is similar to the extraction logic of the upper limit parameter. According to the definition rules of the data structure, the second index node of the one-dimensional floating-point array is located, and the value stored at that memory location is retrieved. This value is generated by normalizing it according to the same standard based on the theoretical minimum value of image background noise, which is the physical absolute value of 0. Dividing 0 by the grayscale range of 255 produces a dimensionless lower limit scaling factor. This factor represents the lowest possible compression depth basis probability that the clipping clipping operation can achieve. This fixed lower limit value is extracted from the corresponding third dimension of the clipping evaluation base vector, thus obtaining... The specific value assigned is 0.0; The parameter acquisition steps are as follows: This parameter represents the local speckle signal-to-noise ratio (SNR) value within the baseline vector for amplitude limiting evaluation. It is obtained by reading the third index position of the one-dimensional array. This value is the true pure proportion feature statistic calculated in the previous sliding window traversal statistics step by dividing the mean and variance parameters of the 25 pixel-level data points within the local region after dimensionless extraction. In the example of the coordinate system in the 10th column horizontally and 10th row vertically of the previous sliding window, its specific ratio was calculated to be 1.2. This ratio is packaged into the vector and passed to this calculation and reading step. The pure proportion value in this dimension is extracted as the core reference for evaluating the purity of the local signal and incorporated into the subsequent adaptive adjustment process. The specific value assigned is 1.2; The steps for obtaining the parameter are as follows: This parameter represents the difference constant between extreme values. It requires retrieving the previously established two-dimensional data table of the local dynamic range parameter set, which has undergone bubble sorting and validity cleaning and filtering. Based on the spatial coordinates of the currently processed pixel, a perfectly matching node row is retrieved, and the corresponding difference constant is read from the second column of that row. The specific value assigned is 0.5; The parameter acquisition steps are as follows: This parameter represents a preset reference grayscale level constant, the value of which is extracted and calculated based on the benchmark calibration data of the probe in a standardized tissue-like ultrasound phantom experiment. 1000 frames of background images are continuously acquired in a test water tank filled with pure coupling agent and a standard scatterer. The average grayscale noise floor physical absolute value of all target-free areas is calculated. For example, if the average noise floor grayscale value of the calibration test summary is 51, this absolute physical value is also divided by the highest dynamic range of 255 for normalization division and forced removal of grayscale units. Dividing 51 by 255 yields 0.2. This calculation result is written into the configuration file as a completely dimensionless smoothing damping ratio parameter for global calculation. This is how the parameter is obtained. The specific value assigned is 0.2; Calculations based on parameters: Substitute the extracted pure proportional parameters into the formula. =1.0, =0.0, =1.2, =0.5, =0.2; Calculate the pure proportional product term of the difference constant between the local speckle signal-to-noise ratio and the extreme values: ; Calculating the square of the above product term yields the dimensionless dynamic weight: ; Calculate the square of the preset reference gray level constant: ; Calculate the sum of the multiplications of the terms in the numerator: ; Calculate the sum of the terms in the denominator: ; Calculate the final proportional quotient when the numerator is divided by the denominator: ; The results show that the dimensionless threshold coefficient calculated by the adaptive truncation limiting variable in the unified probability space is 0.90, which means that the maximum grayscale truncation threshold of the current local region should be set at 90% of the full range of the image. This represents a safe downward adjustment compared to the original physical maximum range limit of 1.0. This is because although the current local region has a high signal-to-noise ratio of 1.2, the difference constant is at a moderate level. The addition of the damping term weakens the extreme tendency of extreme value stretching. This evaluated ratio of 0.90 will be used as the unified scale adjustment standard for the final dynamic limiting operation, and will be used to replace and suppress abnormally bright artifact isolated points that exceed this intensity after denormalization.

[0034] The steps for obtaining the corrected output image matrix are as follows: The original coordinates of the gray level data of the center pixel within the local dynamic range parameter set are located. The extreme points at the original coordinates are replaced and truncated according to the adaptive truncation and limiting variable. The weights are allocated according to the spatial distance ratio and the truncation level values ​​of adjacent histogram regions are cross-fused to establish the corrected output image matrix.

[0035] Specifically, the original coordinates of the center pixel grayscale level data within the local dynamic range parameter set are located. By reading the coordinate mapping index table bound within the parameter set, the focus of the calculation is shifted back to the original two-dimensional amplitude space matrix. For example, the center coordinate point of the 10th column horizontally and the 10th row vertically is located. The original eight-bit grayscale absolute value corresponding to this coordinate point is extracted, and the center maximum extreme point is recorded as the absolute physical grayscale 240. Based on the adaptive truncation and amplitude limiting variable, the extreme point at the original coordinates is replaced and truncated. The previously calculated zero-grayscale value is then used to calculate the zero-grayscale value. The dimensional adaptive truncation limiting variable coefficient 0.90 is restored to absolute physical dimensions. Multiplying 0.90 by the full-scale range of 255 yields an inverse normalized physical truncation threshold of 229.5. Rounding this threshold to 230, the original extreme point data 240 is extracted and compared with this truncation threshold 230. If 240 is greater than 230, a local overexposure signal limiting mechanism is triggered, forcibly erasing the original abnormal grayscale value at that coordinate point and replacing it with 230. This completes the process. Forced level suppression of isolated points of highlight artifacts is achieved by assigning weights based on spatial distance ratios and cross-merging the truncated level values ​​of adjacent histogram regions. Eight adjacent pixels around the current coordinate point are extracted to form a rectangular fusion neighborhood. The Euclidean two-dimensional physical distance from each adjacent grid point to the center coordinate point is calculated, and the reciprocal of the distance is used as the basic calculation standard for fusion weight allocation. The closer the physical distance, the greater the smoothing weight is assigned. For example, adjacent pixels in the four orthogonal directions (up, down, left, right) are uniformly assigned a coefficient weight of 0.15, while pixels in the four diagonal directions are uniformly assigned a coefficient weight of 0.05. The center pixel itself retains a core data weight of 0.20. The truncated level values ​​in these nine pixel regions are multiplied by the corresponding spatial weight ratio parameters and then uniformly summed to smooth the hard edges and mosaic phenomenon of level steps caused by truncation at the boundaries of different image regions. All the new values ​​after extreme value replacement and spatial fusion operations are filled into a brand new blank two-dimensional memory grid to establish a corrected output image matrix.

Claims

1. An ultrasound image correction system based on dynamic range, characterized in that, The system includes: The signal cepstral conversion module is used to extract the ultrasonic image scan line signal from the ultrasonic probe, extract the depth level sequence along the axial depth direction and convert it into logarithmic spectrum level parameters to obtain the cepstral transformation coefficient matrix, extract the amplitude and reciprocal frequency values ​​corresponding to the extreme points of equally spaced isolated peaks in the cepstral transformation coefficient matrix, form the feature vector parameters, and generate the reverberation period peak fraction. The artifact frequency band suppression module is used to extract the amplitude value based on the reverberation period peak fraction, compare the amplitude value with the threshold constant, extract the generated artifact frequency band position coordinates, and set the artifact frequency band position coordinates to zero by associating them with the cepstral domain frequency band coefficient value, thereby establishing an artifact-free spatial echo signal. The local speckle statistics module is used to set the sliding window size parameter according to the artifact-free spatial echo signal, extract the pixel gray level data within the sliding window, calculate the local speckle signal-to-noise ratio value, calculate the difference constant between the extreme values ​​corresponding to the pixel gray level data according to the sliding window size parameter, and obtain the local dynamic range parameter set. The image adaptive correction module is used to calculate an adaptive truncation limiting variable based on the local dynamic range parameter set, extract the original coordinates of the gray level data of the center pixel in the local dynamic range parameter set and the adaptive truncation limiting variable, truncate and assign values ​​to the extreme points at the original coordinates, and establish a corrected output image matrix.

2. The ultrasound image correction system based on dynamic range according to claim 1, characterized in that, The steps for obtaining the peak reverberation period fraction are as follows: The ultrasonic image scan line signal is extracted from the ultrasonic probe. The amplitude value of the ultrasonic image scan line signal is extracted point by point along the axial depth direction. The depth level sequence is constructed according to the order of the amplitude values. Logarithmic calculation is performed on the depth level sequence to convert the depth level sequence into logarithmic spectrum level parameters. Inverse Fourier transform operation is performed on the logarithmic spectrum level parameters in sequence to extract the real part data after transformation and obtain the cepstral transform coefficient matrix. Scan the cepstral transform coefficient matrix along the inverted frequency axis, calculate the difference slope between adjacent values ​​of the cepstral transform coefficient matrix, locate the coordinate point where the difference slope changes from positive to negative and mark it as an isolated peak extreme point, calculate the coordinate difference between adjacent isolated peak extreme points, select points with equal coordinate differences as equally spaced isolated peak extreme points, extract the absolute value of the amplitude and the inverted frequency axis coordinate value of the equally spaced isolated peak extreme points, and obtain the amplitude and inverted frequency value corresponding to the equally spaced isolated peak extreme points; The amplitude and reciprocal frequency values ​​corresponding to the isolated peak extreme points with equal intervals are arranged in ascending order of coordinates. The combined data column constitutes the feature vector parameters. The sum of squares of all values ​​in the feature vector parameters is calculated. The total number of values ​​contained in the feature vector parameters is counted. The sum of squares is divided by the total number of values ​​to obtain the quotient value. The quotient value is extracted as the peak score of the reverberation period.

3. The ultrasound image correction system based on dynamic range according to claim 1, characterized in that, The steps for obtaining the artifact frequency band position coordinates are as follows: The amplitude and reciprocal frequency values ​​within the peak fraction of the reverberation period are analyzed. When the amplitude value is greater than the threshold constant, the distribution trajectory of the reciprocal frequency value and the surface depth parameter of the acoustic probe are compared, the coordinates of the coincidence of the distribution trajectory are extracted, and the position coordinates of the artifact frequency band are generated.

4. The ultrasound image correction system based on dynamic range according to claim 1, characterized in that, The steps for acquiring the artifact-free spatial echo signal are as follows: Based on the artifact frequency band position coordinates, read the cepstral domain frequency band coefficient value, find the index node corresponding to the artifact frequency band position coordinates within the cepstral domain frequency band coefficient value, modify the cepstral domain frequency band coefficient value associated with the index node to zero, and generate a zero-assignment relationship value. The zero-assignment correlation values ​​are spatially transformed and mapped to convert them into a time-dimensional sound wave sequence. The amplitude distribution of the sound wave sequence is extracted, the amplitude space matrix is ​​reconstructed, and an artifact-free spatial echo signal is established.

5. The ultrasound image correction system based on dynamic range according to claim 1, characterized in that, The steps for obtaining the local speckle signal-to-noise ratio value are as follows: Based on the artifact-free spatial echo signal, the length and width boundaries are defined to set the sliding window size parameters. The pixel grayscale level data within the sliding window are extracted by traversing the spatial coordinate system. The mean parameter and variance parameter are statistically analyzed and divided to obtain the local speckle signal-to-noise ratio value.

6. The ultrasound image correction system based on dynamic range according to claim 1, characterized in that, The steps for obtaining the local dynamic range parameter set are as follows: Based on the local speckle signal-to-noise ratio value, the sliding window size parameter and pixel gray level data corresponding to the generation process are retrieved, the maximum and minimum extreme values ​​within the pixel gray level data are located, and the extreme value subtraction difference constant is calculated. Based on the extreme value subtraction difference constant, the corresponding dimension of local speckle signal-to-noise ratio is extracted by callback. The extreme value subtraction difference constant and the local speckle signal-to-noise ratio value are combined to form a set of mapping independent variable parameters. The data columns in the set are arranged to obtain the local dynamic range parameter set.

7. The ultrasound image correction system based on dynamic range according to claim 1, characterized in that, The steps for obtaining the adaptive truncation limiting variable are as follows: Traverse the data nodes of the local dynamic range parameter set, peel off the upper limit and lower limit of the data interval boundary corresponding to the mapping independent variable parameter set, extract the local speckle signal-to-noise ratio value associated with the local dynamic range parameter set, combine the upper limit of the data interval boundary, the lower limit of the data interval boundary and the local speckle signal-to-noise ratio value to generate the amplitude limiting evaluation base vector. Extract the upper limit of the interval boundary, the lower limit of the interval boundary, and the local speckle signal-to-noise ratio value within the baseline vector of the amplitude limiting evaluation. Determine the difference constant between extreme values ​​and calculate the adaptive truncation amplitude limiting variable.

8. The ultrasound image correction system based on dynamic range according to claim 1, characterized in that, The steps for obtaining the corrected output image matrix are as follows: The original coordinates of the gray level data of the center pixel within the local dynamic range parameter set are located. The extreme points at the original coordinates are replaced and truncated according to the adaptive truncation and limiting variable. The weights are allocated according to the spatial distance ratio and the truncation level values ​​of adjacent histogram regions are cross-fused to establish a corrected output image matrix.

Images