Granted Patent B2

US 12,740,768 · App. 18/837,369 · Granted Sep 22, 2026

Inventors: Kun Peng (Nanjing, CN); Feng Yin (Nanjing, CN)

Assignee: Medlander Medical Technology Inc.

A61B8/5253G06T3/4038G06T7/337G06T2207/10132G06T2207/20021G06T2207/20212

Abstract

Claims (42)

1 . A method for ultrasonic panoramic imaging, comprising:

(1) acquiring one frame from to-be-stitched images, determining whether this frame is a first frame among a plurality of the to-be-stitched images, and directly using it as an initial reference image C if it is the first frame or otherwise proceeding to step (2) if it is not the first frame;

(2) acquiring an image A other than the reference image C, expanding a perimeter of the image A based on values of neighboring pixels thereof to generate an image A*, and expanding a perimeter of the reference image C based on values of neighboring pixels thereof to generate an image C*;

(3) dividing the image A* and the image C* equally into n areas, calculating by a sum of absolute differences (SAD) method an average offset of image A* relative to the image C* in each area, where n is a positive integer;

(4) obtaining average offset values in X and Y directions of image coordinates in image A* relative to the image C* for each of the n areas, adding corresponding average offsets to the coordinates of the center points of each area respectively to acquire a new coordinate of n points for the image A relative to the image C, obtaining average offsets of the overall image by averaging the new coordinates of the n points, and obtaining coordinate mapping values for all points of image A relative to the image C after interpolation is performed on the new coordinates of the n points;

(5) traversing each pixel of the image A based on the obtained coordinate mapping values, performing a mapping thereon according to corresponding coordinates and deforming it into a registered image B;

(6) traversing each pixel of the registered image B and the reference image C by calculating a difference of pixels at corresponding positions to obtain a set of pixel differences; traversing the set of pixel differences by comparing the set with 99 respectively and taking the smaller value to obtain a new set of differences;

(7) obtaining a lookup table based on a calibration curve;

(8) correcting the registered image B to obtain a to-be-stitched image B*; and

(9) stitching, by weighted average method for overlapping areas, the to-be-stitched image B* with one previous panoramic resulting image C based on overall average offset, to obtain a composite image D.

2 . The method for ultrasonic panoramic imaging according to claim 1 , wherein step (3) further comprises:

(3.1) setting a square-shaped comparison unit as a minimum comparison range for each area, and in a first comparison operation, an area where the comparison unit is located in upper left corner of A*1 is denoted as Kernel1, and the area where the comparison unit is located in upper left corner of C*1 is denoted as Kernel2;

(3.2) performing subtraction on each pixel point of Kernel1 and Kernel2 to calculate the difference with the pixel values of corresponding points and obtain differences of a plurality of corresponding pixels, then summing up the differences of all the pixels to obtain the Sum of Absolute Differences (SAD) results for Kernel1 and Kernel2 denoted as Sum1, and saving center coordinate common to Kernel1 and Kernel2 denoted as (x1, y1);

(3.3) moving, in a second comparison, Kernel1 and Kernel2 one step to a right side respectively and continuing to repeat step (3.2) to obtain a second SAD result denoted as Sum2 and coordinate (x2, y2), repeating step (3.2) again to further move Kernel1 and Kernel2 one step to the right side until Kernel1 and Kernel2 move to the upper right corner of A*1 and C*1, then moving one step downward to continue with the SAD calculation until Kernel1 and Kernel2 move across the entire areas of A*1 and C*1, and thus obtaining k2 SAD results including k2 Sum values;

(3.4) identifying a minimum value out of said k2 Sum values and a coordinate value (xmin, ymin) corresponding to the minimum value to serve as a final calculation result for this area, and this coordinate value represents the coordinate of the position with the highest similarity in position within the area of the two images; and

(3.5) calculating the average offset value for this area in which coordinate of the center point of the area is subtracted from the coordinate value (xmin, ymin) to obtain the average offsets OffsetX1 and OffsetY1 in X and Y directions.

3 . The method for ultrasonic panoramic imaging according to claim 2 , wherein step (4) further comprises:

calculating for all the areas the average offset values in X and Y directions {OffsetX1, OffsetX2 . . . OffsetXn}, {OffsetY1, OffsetY2 . . . OffsetYn}; and

adding coordinate of the center point of each area in the image A* to corresponding average offsets respectively to obtain new mapping coordinates of n points for the image A relative to the image C, and obtaining average offsets OffsetX, OffsetY of the overall image by averaging the average offset values in the X and Y directions across all the areas.

4 . The method for ultrasonic panoramic imaging according to claim 3 , further comprising: performing an interpolation operation on the new mapping coordinates of the n points for the image A relative to the image C, to obtain the coordinate mapping values for all the points of the image A relative to the image C.

5 . The method for ultrasonic panoramic imaging according to claim 4 , wherein step (5) of traversing each pixel of image A, performing the mapping thereon according to the corresponding coordinates and deforming it into the registered image B is performed with a formula given by

dst[ x,y ]=src[ y Map[ x,y],x Map[ x,y]],

where dst denotes a target image, i.e., the registered image B;

src denotes a source image, i.e., image A; x and y represent x-axis coordinate and y-axis coordinate, respectively; and

yMap and xMap represent y-coordinate mapping and x-coordinate mapping from coordinate mapping values of all the points of image A relative to the image C in step (4).

6 . The method for ultrasonic panoramic imaging according to claim 5 , wherein the calibration curve of step (7) is given by:

OutputValue=Initial Value+(FinalValue−Initial Value)/(1+exp(−Slope*4*(InputValue−Brkpt))),

where InitialValue denotes an initial value of the curve;

Final Value denotes a final value of the curve;

Slope is a slope of the curve;

Brkpt denotes a value of inflection point of the curve;

InputValue denotes input parameter, which is a collection consisting of a plurality of numerical values; and

OutputValue denotes a set of output values obtained by substituting the values of, which is referred to as the lookup table and is denoted as the arrLUT dataset.

7 . The method for ultrasonic panoramic imaging according to claim 6 , wherein step (8) further comprises:

obtaining a set of registration coefficient Coeffx*y in accordance with a formula given by Coeff[i]=arrLUT[Diff*[i]] where i∈[0,x*y), and

correcting the registered image B in accordance with a formula given by ImageB[i]=Coeff[i]*ImageB[i]+(1−Coeff[i])*ImageC[i], where i∈[0, x*y), ImageB[i] denotes i-th pixel of the image B, ImageC[i] denotes i-th pixel of image C, resulting in the to-be-stitched image B*.

8 . The method for ultrasonic panoramic imaging according to claim 7 , wherein the stitching by the weighted average method in step (9) is specified by:

(9.1) stitching the to-be-stitched image B* onto the reference image C based on the overall average offset OffsetX and OffsetY, to obtain the composite image D, wherein there is an overlapping area between the two images in the composite image D which requires further correction and smoothing;

(9.2) setting an overlap coefficient SpaceCoeffCount, indicating that image correction is performed for the area having a width defined by SpaceCoeffCount in the overlapping area, and obtaining a correction dataset arrCoefficient with a number of SpaceCoeffCount in accordance with a formula given by arrCoefficient[i]=((i+1))/SpaceCoeffCount where i∈[0, SpaceCoeffCount);

(9.3) Basing on a formula given by ImageD[x,y]=(1−arrCoefficient[x])*ImageC[x,y]+arrCoefficient[x]*ImageB*[x,y], where [x,y] denotes coordinate of a point in the image, ImageD denotes the composite image D, ImageC denotes the reference image C, and ImageB* denotes the to-be-stitched image B*.

9 . A computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor is configured to execute the computer program to implement the method according to claim 1 .

10 . A non-transitory computer-readable storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, is configured to implement the method according to claim 1 .

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (35)