Redness differentiation calibration (RDC) method for medical image

Abstract

The present invention relates to the technical field of image processing, and specifically relates to a redness differentiation calibration (RDC) method for a medical image, the method comprising the following steps: acquiring an original image, and obtaining values L, a and b of the original image; performing redness intensity estimation on the original image by means of the values a and b; performing edge information extraction on the original image by means of the value L; calibrating the original image by means of calculation, so as to obtain a calibration result LQ; and using the obtained LQ and the obtained values a and b to convert a CIELAB color space of the original image into an RGB color space, and outputting a result image. Compared with a conventional method, the present technical solution can effectively estimate the redness intensity, calibrates a mucosal structure and tissue features, and highlights microvessels in a superficial mucosa and a submucosa in an endoscope image; moreover, transformation on a non-red area is less, such that a processed image is more suitable for a doctor to observe and make a diagnosis.

Description

[Rule 91 correction 08.04.2025] A medical image redness differentiation calibration method (R.D.C., Redness Differentiation Calibration) TECHNICAL FIELD

[0001] [Rule 91 correction 08.04.2025] The present application relates to the technical field of image processing, in particular to a medical image redness differentiation calibration method (R.D.C., Redness Differentiation Calibration). BACKGROUND

[0002] Modern endoscopes play a key role in various gastrointestinal-related disease diagnosis, endoscopic surgery and other scenarios, and good endoscopic image quality helps improve diagnostic levels.

[0003] Currently, due to the non-obvious spectral differentiation of the reflected spectrum of mucosa, blood vessels and tissues in the red spectral band, the endoscopic camera system does not show enough red color levels in key scenes such as examination and surgery, which brings inconvenience to doctors in the treatment and judgment process. SUMMARY

[0004] [Rule 91 correction 08.04.2025] In order to overcome the above technical problems, the purpose of the present application is to provide a medical image redness differentiation calibration method (R.D.C., Redness Differentiation Calibration) to improve the red color levels shown by the endoscope in key scenes such as examination and surgery.

[0005] The purpose of the present application can be achieved by the following technical solutions:

[0006] [Rule 91 correction 08.04.2025] A medical image redness differentiation calibration method (R.D.C., Redness Differentiation Calibration) comprises the following steps:

[0007] S1, obtaining an original image, converting the RGB color space of the original image into a CIELAB color space to obtain L, a, b values of the original image;

[0008] S2, using the a and b values obtained in step S1 to estimate the red intensity of the original image;

[0009] S3, using the L value obtained in step S1 to extract edge information of the original image;

[0010] S4, using the L value, red intensity result and edge information result obtained by S1 step, S2 step and S3 step, enhancing the original image by calculation to obtain an enhanced result L Q ;

[0011] S5, using the obtained L Q , a and b values, converting the CIELAB color space of the original image into RGB color space, and outputting a result image.

[0012] Further, in S1 step, the color depth of the original image to be processed at least includes 8-bit image and 10-bit image.

[0013] Further, in S1 step, the specific method for calculating L, a, b values is:

[0014] S11, converting the RGB range of the original image to 0-1 to obtain target R, G, B values;

[0015] S12, converting the obtained target R, G, B values into X, Y, Z values by the following formula, and then converting the X, Y, Z values into L, a, b three channel component values in CIELAB color space, wherein the range of L, a, b three channel component values is 0-255:

[0016] obtaining X, Y, Z values; a = 500(f(X)-f(Y))+256#(4); b = 200(f(Y)-f(Z))+256#(5);

[0017] In formula (4), (5):

[0018] Finally, L, a, b three channel component values are obtained.

[0019] Further, after obtaining X, Y, Z values by formula (1), the obtained X, Y, Z values are normalized according to the X, Y, Z values specified by the reference white point D65:

[0020] wherein X n = 0.950456;

[0021] wherein Y n = 1;

[0022] wherein Z n = 0.950456.

[0023] Furthermore, in step S2, the specific method for estimating the red intensity of the original image is:

[0024] S21. Perform the following calculation using the obtained a and b values: redness=(weightA*weightB) γ #(12);

[0025] Among them, abs() means taking the absolute value, and γ is the weight adjustment coefficient;

[0026] S22. Normalize the obtained result to obtain the estimated red intensity:

[0027] Among them, max() means taking the maximum value.

[0028] Furthermore, in step S3, the specific method for extracting edge information from the original image is: edge=LL blur #(14);

[0029] Among them, L blur Represents the blurred image of the L channel.

[0030] Furthermore, the blurred image of the L channel is achieved by using a 9*9 Gaussian filter.

[0031] Furthermore, in step S4, the specific method of enhancing the original image by calculation is:

[0032] S41. First, the L channel is transformed using the following nonlinear mapping function:

[0033] Among them, e, f, g, and h are the adjustment parameters of the nonlinear mapping function, which adjust the inflection point and stretching degree of the mapping curve;

[0034] S42. Adjust the intensity of image enhancement using the following formula: DeltaL = (f(L) - L) * coeffL + edge * coeffE# (16);

[0035] Among them, coeffL=5, coeffE=4.5;

[0036] S43, the enhancement result L is obtained by the following formula Q , where L Q The range is between 0-255: L Q =L+DeltaL*weight#(17).

[0037] Furthermore, e=5, f=1.7, g=8, and h=1.4.

[0038] Further in the step S5, in the conversion, firstly the X, Y, Z values to be converted are obtained by the following formula: X=0.950456*x r (18); Y=y r (19); Z=1.088754*z r (20);

[0039] Wherein,

[0040] Then the obtained X, Y, Z values to be converted are filled in the following formula, and the r, g, b values are obtained:

[0041] Finally, the r, g, b values are restored to the R, G, B range of the original image, and the result image is output.

[0042] The beneficial effects of the present application are:

[0043] Compared with the traditional method, the technical solution converts the RGB color space of the original image into the CIELAB color space, obtains the L, a, b values of the original image, then estimates the red intensity of the original image and extracts the edge information in turn, and enhances the original image by calculation, obtains the enhanced result L Q , and finally converts the CIELAB color space of the original image into the RGB color space and outputs the result image, and the results show that the method estimates the red intensity of each pixel point of the original image, enhances the contrast of the red region, improves the color differentiation degree and clarity of the key parts of the edge, strengthens the display of the blood vessels and structure of the surface mucosa, and has important significance in the discovery and diagnosis of micro lesions, targeted biopsy, positioning of lesion boundary, tissue positioning in the surgical process and the like. BRIEF DESCRIPTION OF DRAWINGS

[0044] The present application will be further described below in conjunction with the drawings.

[0045] [According to Rule 91, correct on 08.04.2025] Fig. 1 is a processing flow diagram of the medical image red differentiation enhancement method (R.D.C., Redness Differentiation Calibration) proposed by the present application;

[0046] Fig. 2 is an image example of an original endoscope;

[0047] Fig. 3 is an output image of the image in Fig. 2 after red intensity estimation;

[0048] Fig. 4 is an output image of the image in Fig. 3 after edge information extraction;

[0049] Fig. 5 is an output image after the image in Fig. 4 is calculated and enhanced;

[0050] Fig. 6 is an output image after the image in Fig. 5 is converted in the LAB color space. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0052] [Corrected according to Rule 91 on 08.04.2025] As shown in Figs. 1, 2, 3, 4, 5, and 6, a medical image redness differentiation calibration method (R.D.C., Redness Differentiation Calibration) comprises the following steps:

[0053] S1, an original image is obtained, and the RGB color space of the original image is converted into the CIELAB color space to obtain the L, a, and b values of the original image. The CIELAB is a device-independent color space, which includes three channels of brightness L, a range from red to green a, and a range from yellow to blue b, can better describe the perceived properties of colors, and can more accurately represent and process the color information of the image by converting the RGB color space to the CIELAB color space, which is conducive to color correction, color matching, image segmentation, and other color-related image processing tasks.

[0054] The color depth of the original image processed by the method includes 8-bit images and 10-bit images;

[0055] S11, when the original image to be processed is an 8-bit image, it is necessary to first convert the RGB range of the original image to between 0 and 1, which can be achieved by dividing the R, G, and B values to be input in the original image by 255, and finally obtaining the target R, G, and B values;

[0056] S12, the target R, G, and B values obtained are converted into X, Y, and Z values by the following formula, and the X, Y, and Z values are converted into L, a, and b three channel component values in the CIELAB color space, wherein the range of the L, a, and b three channel component values is between 0 and 255:

[0057] After X, Y, Z values are obtained by formula (1), the obtained X, Y, Z values are normalized according to the X, Y, Z values defined by reference white point D65:

[0058] wherein X n = 0.950456;

[0059] wherein Y n = 1;

[0060] wherein Z n = 0.950456.

[0061] After the normalized X, Y, Z values obtained according to formula (7), (8), (9), the L, a, b three channel component values can be obtained by using the following formula: a = 500 (f (X) - f (Y) ) + 256 # (4) ; b = 200 (f (Y) - f (Z) ) + 256 # (5) ;

[0062] In formula (4), (5):

[0063] Finally, the L, a, b three channel component values are obtained.

[0064] S2, as shown in FIG. 2, FIG. 3, using the a and b values obtained by S1 step, the red intensity of the original image is estimated, and the specific method is:

[0065] S21, the following calculation is performed by using the obtained a, b values: redness = (weightA * weightB) γ # (12) ;

[0066] Wherein, abs () represents taking absolute value, and γ is a weight adjustment coefficient, which is 2 in the embodiment;

[0067] S22, the obtained result is normalized to obtain the estimated red intensity:

[0068] Wherein, max () represents taking maximum value.

[0069] S3, as shown in FIG. 2, FIG. 4, using the L value obtained by S1 step, the edge information of the original image is extracted.

[0070] The common edge information extraction method is to subtract the blurred image from the original gray scale image, so as to obtain the high frequency component in the gray scale image, and in the method, the L channel component value is subtracted from the L channel blurred image, that is: edge = L-Lblur (14) ;

[0071] wherein, L blur represents the blurred image of L channel, the blurring method of the blurred image of L channel is not limited, in the embodiment, the blurred image of L channel is realized by using a 9*9 Gaussian filter.

[0072] S4, as shown in FIG. 2, FIG. 3, FIG. 4, FIG. 5, the L value, the red intensity result and the edge information result obtained by using the S1 step, the S2 step and the S3 step are used to enhance the original image by calculation, and the enhanced result L Q is obtained. Through enhancement, the quality of the image can be improved, the specific features of the image can be enhanced, or the visualization effect of the image can be improved, so that the image is clearer, easier to analyze and has higher contrast. The specific method of calculation is as follows:

[0073] S41, first, the L channel is transformed by using the following nonlinear mapping function:

[0074] wherein, e, f, g, h are adjustment parameters of the nonlinear mapping function, which can adjust the inflection point and stretching degree of the mapping curve. In the embodiment, e = 5, f = 1.7, g = 8, and h = 1.4.

[0075] S42, the intensity of image enhancement is adjusted by the following formula: DeltaL = (f(L) - L) * coeffL + edge * coeffE (16) ;

[0076] wherein, coeffL = 5, and coeffE = 4.5.

[0077] S43, the enhanced result L Q is obtained by the following formula: Q wherein, L Q ranges between 0 and 255: L Q = L + DeltaL * weight (17).

[0078] S5, as shown in FIG. 6, the CIELAB color space of the original image is converted into the RGB color space by using the obtained L r , a and b values, and the result image is output:

[0079] In the conversion, first, the X, Y and Z values to be converted are obtained by the following formulas: X = 0.950456 * x r (18) ; Y = y r (19) ; Z = 1.088754 * z r (20) ;

[0080] wherein,

[0081] Then the obtained X, Y, Z values to be converted are filled into the following formula to obtain r, g, b values:

[0082] Finally, the r, g, b values are restored to the R, G, B range of the original image, and the result image is output.

[0083] The scheme converts the RGB color space of the original image into the CIELAB color space to obtain the L, a, b values of the original image, then estimates the red intensity and extracts the edge information of the original image in turn, and enhances the original image by calculation to obtain the enhanced result L Q Finally, the CIELAB color space of the original image is converted into the RGB color space, and the result image is output. The results show that the method can effectively estimate the red intensity, enhance the mucosa structure and tissue characteristics, and highlight the microvessels in the superficial layer of the mucosa and the submucosa in the endoscope image, while the transformation of the non-red area is less, so that the processed image is more suitable for doctors to observe and make a diagnosis.

[0084] In the description of the present specification, the description of the terms "one embodiment", "example", "specific example" and the like means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are contained in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0085] The above is only an example and description of the present application, and those skilled in the art can make various modifications or supplements to the described specific embodiments or use similar ways to replace, as long as they do not deviate from the invention or exceed the scope defined by the present claims, which shall belong to the protection scope of the present application.

Claims

1. [Corrected 08.04.2025 according to Rule 91] A method for redness differentiation calibration (RDC) of medical images, characterized by: The following steps are involved: S1. Obtain the original image, convert the RGB color space of the original image into the CIELAB color space, and obtain the L, a, and b values ​​of the original image; In step S1, the specific method for calculating the L, a, and b values ​​is: S11, converting the RGB range of the original image to between 0 and 1 to obtain the target R, G, and B values; S12. Convert the obtained target R, G, and B values ​​to X, Y, and Z values ​​using the following formula, and then convert the X, Y, and Z values ​​to L, a, and b channel component values ​​in the CIELAB color space, where the L, a, and b channel component values ​​range from 0 to 255: Get the X, Y, and Z values; a=500(f(X)-f(Y))+256#(4); b=200(f(Y)-f(Z))+256#(5); In formulas (4) and (5): Finally, the three channel component values ​​of L, a, and b are obtained; After obtaining the X, Y, and Z values ​​through formula (1), the obtained X, Y, and Z values ​​are normalized according to the X, Y, and Z values ​​specified by the reference white point D65: where X n =0.950456; where Y n =1; where Z n =0.950456; S2, using the a and b values ​​obtained in step S1, estimate the red intensity of the original image; In step S2, the specific method for estimating the red intensity of the original image is: S21. Perform the following calculation using the obtained a and b values: redness=(weightA*weightB) γ #(12); Among them, abs() means taking the absolute value, and γ is the weight adjustment coefficient; S22. Normalize the obtained result to obtain the estimated red intensity: Among them, max() means taking the maximum value; S3, using the L value obtained in step S1 to extract edge information from the original image; In step S3, the specific method for extracting edge information from the original image is: edge = LL blur #(14); Among them, L blur represents the blurred image of L channel; S4, using the L value, red intensity result and edge information result obtained in step S1, step S2 and step S3, the original image is enhanced by calculation to obtain the enhanced result L Q ; In step S4, the specific method of enhancing the original image by calculation is: S41. First, the L channel is transformed using the following nonlinear mapping function: Among them, e, f, g, and h are the adjustment parameters of the nonlinear mapping function, which adjust the inflection point and stretching degree of the mapping curve; S42. Adjust the intensity of image enhancement using the following formula: DeltaL = (f(L) - L) * coeffL + edge * coeffE# (16); Among them, coeffL=5, coeffE=4.5; S43, the enhancement result L is obtained by the following formula Q , where L Q The range is between 0-255; L Q =L+DeltaL*weight#(17); S5. Apply the obtained L Q , a, and b values, converts the CIELAB color space of the original image to RGB color space, and outputs the resulting image.

2. [Corrected 08.04.2025 according to Rule 91] A method for redness differentiation calibration (RDC) of medical images according to claim 1, characterized in that: In step S1, the color depth of the original image to be processed includes at least 8-bit image and 10-bit image.

3. [Corrected 08.04.2025 according to Rule 91] A method for redness differentiation calibration (RDC) of medical images according to claim 1, characterized in that: The blurred image of the L channel is achieved by using a 9*9 Gaussian filter.

4. [Corrected 08.04.2025 according to Rule 91] A method for redness differentiation calibration (RDC) of medical images according to claim 1, characterized in that: e=5, f=1.7, g=8, h=1.

4.

5. [Corrected 08.04.2025 according to Rule 91] A method for enhancing redness differentiation of medical images (RDC) according to claim 4, characterized in that: In step S5, when converting, first obtain the X, Y, and Z values ​​to be converted using the following formula: X = 0.950456*x r #(18); Y=y r #(19); Z = 1.088754*z r #(20); in, Then fill the obtained X, Y, and Z values ​​to be converted into the following formula to obtain the r, g, and b values: Finally, the r, g, and b values ​​are restored to the R, G, and B ranges of the original image, and the resulting image is output.

Images