Intelligent fluorescence microscopic image enhancement method and device, and readable storage medium

Abstract

The application discloses a kind of fluorescent microscopic image intelligent enhancement method, device and readable storage medium, belong to medical image processing field.The purpose is to solve the problem of fluorescent image multi-channel fusion difficulty, pseudo-color mapping inaccuracy, RGB processing unevenness and parameter dependence experience.The method comprises: obtaining multi-channel image and separating;Perform pseudo-color mapping and synthesis based on wavelength physical characteristics, to obtain a composite image;Obtain the R, G, B three-channel target histogram features of standard library;The histogram specification of R, G, B three-channel independent of composite image is carried out;Using genetic algorithm, with target histogram similarity as target, adaptively optimize R, G, B three-channel independent Gamma parameter;The optimal parameter is applied to independent Gamma correction, to obtain final enhanced image.The application realizes physical accurate mapping, standardization correction and parameter adaptation, and improves image quality.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medical image processing and computer vision, in particular to an intelligent processing and enhancement method of fluorescence microscopic images. BACKGROUND

[0002] Fluorescence microscopic imaging technology plays a crucial role in biomedical research, pathological diagnosis, and drug screening. However, in the actual process of fluorescence image acquisition and processing, existing technologies still face a series of key challenges.

[0003] Firstly, modern fluorescence microscopes usually collect multiple (e.g., 2-6) fluorescence channels simultaneously, and the emission wavelength and intensity characteristics of different fluorescent probes (such as DAPI, FITC, TRITC, etc.) differ significantly. Traditional image processing methods are difficult to achieve scientific and accurate multi-channel information fusion and visualization.

[0004] Secondly, existing pseudo-color mapping methods mostly rely on simple RGB color mapping or lookup tables (LUT), without fully considering the physical optical characteristics of fluorescence wavelengths, resulting in generated pseudo-color images that do not match the actual fluorescence spectral characteristics, and thus affecting the accuracy and reproducibility of scientific research.

[0005] In addition, traditional image enhancement methods often use uniform processing strategies for the RGB three channels, which ignores the differentiated characteristics and different importance of red, green, and blue channels in fluorescence images, and easily leads to information loss or over-enhancement of some channels. At the same time, the quality assessment and enhancement of fluorescence images generally lack objective standard references, and the quality of images collected by different laboratories or devices is uneven, making it difficult to achieve standardized and normalized processing of image quality.

[0006] Finally, in the image correction (such as Gamma correction) process, the setting of related parameters (such as Gamma value) in traditional methods highly depends on human experience, lacks adaptive optimization mechanism, and is difficult to automatically determine the optimal parameters for different quality input images and different standard references. SUMMARY

[0007] The embodiments of the present application provide a fluorescence microscopic image intelligent enhancement method, device and readable storage medium, which aims at the defects of existing fluorescence microscopic image processing technology, such as difficulty in multi-channel information fusion, lack of physical accuracy in pseudo-color mapping, imbalance in RGB channel processing, lack of standardized reference, dependence on manual experience setting of enhancement parameters, and lack of adaptive optimization.

[0008] The core technology of the present application mainly comprises the following steps: firstly, constructing a pseudo-color mapping algorithm based on the physical optical characteristics of fluorescence wavelength and the visual perception characteristics of human eyes, and generating a composite image by using a maximum synthesis strategy; then, introducing a standard image library to normalize the RGB three channels of the composite image; finally, using a genetic algorithm to adaptively and globally optimize the independent Gamma correction parameters of the RGB three channels, so as to realize the intelligent enhancement of the image.

[0009] In a first aspect, the present application provides an intelligent enhancement method for fluorescence microscopic images, which comprises the following steps:

[0010] Obtaining a multi-channel fluorescence microscopic image and separating it into a plurality of single-channel gray images; establishing the corresponding fluorescence emission wavelengths of the plurality of single-channel gray images;

[0011] Converting the plurality of single-channel gray images into corresponding single-channel pseudo-color images based on the physical optical characteristics of the fluorescence emission wavelengths;

[0012] Synthesizing the plurality of single-channel pseudo-color images to generate a multi-channel pseudo-color composite image;

[0013] Obtaining the target histogram features of the R, G and B three channels of a standard reference image library;

[0014] Respectively performing independent histogram normalization processing on the R, G and B three channels of the multi-channel pseudo-color composite image according to the target histogram features to obtain intermediate correction images;

[0015] Through an optimization algorithm, taking the similarity between the histogram of the intermediate correction image after Gamma correction and the target histogram features as the optimization target, the optimal Gamma correction parameters corresponding to the R, G and B three channels are adaptively searched and determined;

[0016] Applying the optimal Gamma correction parameters to independently perform Gamma correction on the R, G and B three channels of the intermediate correction image to obtain a final enhanced image.

[0017] Further, the step of establishing the fluorescence emission wavelengths comprises:

[0018] According to the detected number of channels, a preset wavelength configuration strategy or a wavelength uniform distribution strategy is used to assign an emission wavelength to each single-channel gray image.

[0019] Further, the conversion step based on the physical optical characteristics comprises:

[0020] Defining a mapping function of the wavelength covering the ultraviolet to near-infrared band to the R, G and B three stimulus values, and combining the wavelength-dependent transparency factor simulating the sensitivity of human eyes to calculate the single-channel pseudo-color image.

[0021] Further, the synthesis step adopts a maximum value synthesis strategy, and maximum values of pixel values of the multiple single-channel pseudo-color images on the R, G and B channels are taken respectively to generate a multi-channel pseudo-color synthesis image.

[0022] Further, the optimization algorithm is a genetic algorithm, and the step of adaptively searching and determining the optimal Gamma correction parameter comprises:

[0023] combining and encoding the Gamma correction parameters of the R, G and B channels into an individual of the genetic algorithm;

[0024] designing a fitness function for evaluating a comprehensive similarity degree between R, G and B channel histograms of the intermediate correction image after the corresponding Gamma correction parameter of the individual is applied to the intermediate correction image and a target histogram feature;

[0025] evolving the population through selection, crossover and mutation operations until convergence or a maximum iteration number is reached to obtain the optimal Gamma correction parameter.

[0026] Further, the fitness function comprehensively considers at least one of the Barth coefficient, the Pearson correlation coefficient and the chi-square distance to measure the comprehensive similarity degree, and the fitness function assigns different weights to the similarity degrees of the R, G and B channels when calculating the comprehensive similarity degree, and the weights are determined based on human visual perception characteristics.

[0027] Further, the step of applying the optimal Gamma correction parameter to perform independent Gamma correction comprises:

[0028] pre-constructing a lookup table for each of the R, G and B channels based on the respective optimal Gamma correction parameter thereof;

[0029] mapping pixel values of the R, G and B channels of the intermediate correction image by using the lookup table to complete the Gamma correction.

[0030] In a second aspect, the present application provides a fluorescent microscopic image intelligent enhancement device, comprising:

[0031] a separation module configured to acquire a multi-channel fluorescent microscopic image and separate the multi-channel fluorescent microscopic image into multiple single-channel grayscale images;

[0032] a wavelength mapping module configured to establish a fluorescent emission wavelength corresponding to each of the multiple single-channel grayscale images;

[0033] a pseudo-color conversion module configured to convert each of the multiple single-channel grayscale images into a corresponding single-channel pseudo-color image based on physical optical characteristics of the fluorescent emission wavelength;

[0034] An image synthesis module is configured to synthesize a plurality of single-channel pseudo-color images to generate a multi-channel pseudo-color composite image.

[0035] A standard feature acquisition module is configured to acquire target histogram features of R, G and B channels of a standard reference image library.

[0036] A histogram specification module is configured to perform independent histogram specification processing on R, G and B channels of the multi-channel pseudo-color composite image according to the target histogram features respectively to obtain an intermediate correction image.

[0037] A parameter optimization module is configured to search and determine optimal Gamma correction parameters of R, G and B channels respectively by an optimization algorithm, taking the similarity between the image histogram of the intermediate correction image after Gamma correction and the target histogram features as an optimization target.

[0038] A final correction module is configured to perform independent Gamma correction on R, G and B channels of the intermediate correction image by applying the optimal Gamma correction parameters of R, G and B channels respectively to obtain a final enhanced image.

[0039] In a third aspect, the present application provides an electronic device comprising a memory and a processor, the memory storing a computer program, and the processor being configured to run the computer program to execute the above-mentioned fluorescent microscopy image intelligent enhancement method.

[0040] In a fourth aspect, the present application provides a readable storage medium, the readable storage medium storing a computer program, the computer program comprising program codes for controlling a process to execute the process, and the process comprising the above-mentioned fluorescent microscopy image intelligent enhancement method.

[0041] The main contributions and innovations of the present application are as follows:

[0042] 1. Physical accuracy and repeatability of pseudo-color mapping: The present application adopts a pseudo-color generation algorithm based on physical optics, covering the full wavelength range from ultraviolet to near-infrared (300-850nm), and considering the wavelength-dependent transparency factor to simulate human eye sensitivity, ensuring that the same wavelength always produces the same color, and the generated pseudo-color image is highly consistent with the actual spectral characteristics, significantly improving the repeatability of scientific research.

[0043] 2. Fidelity and equalization processing of multi-channel information: Through the RGB three-channel independent optimization framework and combined with the maximum value synthesis strategy, the present application effectively solves the channel information loss or over-enhancement problem caused by the traditional unified processing strategy, ensuring that the signals of different fluorescent probes (especially in the red and green channels) are retained and optimized.

[0044] 3. Standardization and adaptability of enhancement processing: the present application solves the problem of lack of objective standard reference of enhancement processing by introducing standard image library for RGB three-channel independent histogram specification; at the same time, the global search of optimal multi-objective fitness function of genetic algorithm overcomes the defects of traditional Gamma correction parameters depending on artificial experience, and realizes adaptive intelligent enhancement of different input images.

[0045] 4. High automation and intelligence of processing flow: the present application realizes the whole flow processing from automatic detection of the number of fluorescence channels of DICOM images, adaptive wavelength configuration to parameter automatic optimization, improves the efficiency and stability of image processing.

[0046] The details of one or more embodiments of the present application are presented in the following drawings and description to make other features, objects and advantages of the present application more clear and easy to understand. BRIEF DESCRIPTION OF DRAWINGS

[0047] The drawings described herein are intended to provide further understanding of the present application, and form a part of the present application. The illustrative embodiments of the present application and their description serve to explain the present application, and do not constitute an improper limitation of the present application. In the drawings:

[0048] Figure 1 is a flow chart of the intelligent enhancement method of fluorescence microscopic images according to the embodiment of the present application;

[0049] Figure 2 is an architectural diagram of multi-channel decomposition and pseudo-color synthesis according to the embodiment of the present application;

[0050] Figure 3 is a detailed diagram of genetic algorithm optimization process according to the embodiment of the present application;

[0051] Figure 4 is a comparison diagram of actual enhancement effect according to the embodiment of the present application;

[0052] Figure 5 is a schematic diagram of the hardware structure of the electronic device according to the embodiment of the present application. DETAILED DESCRIPTION

[0053] The exemplary embodiments will be described in detail herein with reference to the accompanying drawings. Unless otherwise indicated, the same numbers on different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all the implementations consistent with one or more embodiments of the present specification. Instead, they are merely examples of devices and methods consistent with some aspects of one or more embodiments of the present specification as detailed in the appended claims.

[0054] ​It should be noted that the steps of the respective methods are not necessarily performed in the order shown and described in this specification in other embodiments. In some other embodiments, the steps included in its method can be more or less than described in this specification. In addition, a single step described in this specification may, in other embodiments, be broken down into multiple steps for description; and multiple steps described in this specification may, in other embodiments, be combined into a single step for description.

[0055] Embodiment I

[0056] The present application aims to provide an intelligent enhancement method for fluorescence microscopic images. Specifically, with reference to Figure 1 , the method comprises the following steps:

[0057] Step S1: Intelligent reading and preprocessing of multi-channel fluorescence images (corresponding to multi-channel fluorescence image reading in Figure 1 )

[0058] In this embodiment, this step receives fluorescence microscopic images in DICOM format. First, the system automatically detects the hierarchical structure of the input image and the number of fluorescence channels , which is usually in the range of 1 to 8 channels. Subsequently, the image is separated by channel, separating the input multi-dimensional data and extracting each channel corresponding single-channel grayscale image . .

[0059] For subsequent processing, the extracted grayscale images need to be standardized in data type

[0060] Image data structure definition: the input image is represented as a four-dimensional tensor:

[0061]

[0062] In the above formula, represents the real number field, B is the batch size, C is the number of channels (dynamically detected, ranging from 1 to 8), H and W are the image height and width.

[0063] Single-channel grayscale data extraction: for the i-th fluorescence channel , the grayscale image is extracted.

[0064]

[0065] Next, normalize the input data of any type, such as the grayscale image :

[0066]

[0067] In the above formula, and are the global minimum and maximum values of the channel grayscale image, respectively. Then, the normalized floating-point values in the range [0, 1] are mapped to the range [0, 255] and rounded down to convert to unsigned 8-bit integers:

[0068]

[0069] In the above formula, denotes the rounding down operation.

[0070] Step S2: Adaptive fluorescence channel wavelength mapping (corresponding to adaptive fluorescence channel wavelength mapping in Figure 1 )

[0071] Please understand in combination with Figure 2 . The core of this step is to establish the mapping relationship of channel index i to the actual fluorescence emission wavelength . The present application provides an intelligent wavelength mapping strategy.

[0072] Define a double mapping function:

[0073]

[0074] In the above formula, denotes the fluorescence emission wavelength (unit: nm) corresponding to the th channel.

[0075] In a preferred embodiment, for a common number of channels (e.g. 1 to 6 channels), the system adopts a preset optimal wavelength configuration strategy. This strategy is based on examples of commonly used fluorescent probes in biomedical experiments, such as Table 1 below:

[0076] Table 1

[0077]

[0078] In another embodiment, when the number of channels or is a non-standard configuration (number of channels), a general wavelength uniform distribution strategy is adopted:

[0079]

[0080] This strategy ensures coverage of the entire visible to near-infrared wavelength band (350-780 nm).

[0081] Finally, a complete mapping table is established for each channel:

[0082]

[0083] wherein, standard fluorescence probe name (e.g. ", ", etc. ), pseudo-color display color (e.g. " ).

[0084] Step S3: Pseudo-color mapping algorithm based on wavelength physical properties

[0085] This step is the key to realize physically accurate pseudo-color. The present invention defines an algorithm to calculate the corresponding RGB tristimulus values based on human eye visual perception properties and spectral physical properties for a given wavelength . The algorithm preferably covers the full wavelength range from ultraviolet (UV) to near infrared (NIR), for example .

[0086] In a preferred embodiment, the algorithm establishes the mapping relationship between wavelength and normalized red component , green component and blue component by defining piecewise functions. For example:

[0087] Define wavelength-dependent normalized red component :

[0088]

[0089] Define wavelength-dependent normalized green component :

[0090]

[0091] Define wavelength-dependent normalized blue component :

[0092]

[0093] Further, to simulate the difference in sensitivity of the human eye to different wavelengths, the present invention introduces a wavelength-dependent transparency factor :

[0094]

[0095] This factor is designed to have the highest transparency ( ) in the most sensitive visible light wavelength range of the human eye (e.g. ), and gradually attenuates in the UV (e.g. ) and NIR (e.g. ) wavelength range, simulating the decrease in human eye sensitivity.

[0096] Finally, the actual RGB pixel values (0~255), G(0~255) and B(0~255) are calculated by For example:

[0097]

[0098]

[0099] In the above formula, is the maximum pixel intensity of the 8-bit image, represents the floor function.

[0100] It also includes Alpha channel and luminance value calculation:

[0101]

[0102] In the above formula, represents the Alpha channel value, which represents the transparency component of the pseudo-color image. is the perceived luminance, which adopts the luminance weight of standard.

[0103] To improve the calculation efficiency, the LRU cache mechanism is preferred in this embodiment:

[0104]

[0105] Take the hash value of the wavelength (and the possible Gamma value ) as the cache key, and cache the calculated (R, G, B, A, Y) quintuple as the value , cache_value represents the physical accurate pseudo-color value corresponding to the wavelength, avoiding repeated calculation for the same wavelength. Step S4: Multi-channel grayscale to pseudo-color conversion and synthesis (corresponding to pseudo-color mapping and multi-channel synthesis in

[0106] ) Figure 1 Please refer to

[0107] , which shows the multi-channel decomposition and pseudo-color synthesis architecture. This step converts each standardized single-channel grayscale image Figure 2 (grayscale image in ) obtained in S1 into a pseudo-color image. Figure 2

[0108] First, the single-channel grayscale image is expanded to a three-channel representation (that is, the values of the three channels are all . For example:​

[0109]

[0110] i.e.

[0111]

[0112] In the above equation, represents the three-channel expanded image, represents repeating 3 times and outputting a vector of length 3.

[0113] Then, according to the wavelength corresponding to the channel in S2 (wavelength mapping in S2), the corresponding BGR color factor vector is calculated using the algorithm of S3: Figure 2

[0114] In the above equation,

[0115] represents the BGR color factor vector, which is a color factor vector calculated based on the wavelength of the i-th channel .

[0116] Next, the pseudo-color image of the channel (pseudo-color image in S4) is generated by element-wise multiplication (Hadamard product): Figure 2

[0117] In the above equation,

[0118] represents the single-channel pseudo-color image, represents element-wise multiplication, and its expanded form is as follows:

[0119]

[0120] After all single-channel pseudo-color images

[0121] are generated, they need to be combined into a multi-channel pseudo-color composite image (BGR pseudo-color composite image in S5). Figure 2 In a preferred embodiment, the present application adopts a maximum synthesis strategy (multi-channel maximum synthesis in S5). This strategy takes the maximum value on each color channel of each pixel (x, y) for all pseudo-color images of the channels (

[0122] , where N is the total number of channels). Figure 2 ​​​​​​

[0123]

[0124] This strategy effectively preserves the brightest fluorescence signal at each pixel location while avoiding oversaturation caused by traditional additive synthesis.

[0125] Finally, the synthesized image Perform boundary truncation (using the `clip` function) to ensure all pixel values ​​are within the range [0, 255] and convert them to integers. For example:

[0126]

[0127] in, This represents the final synthesized pseudo-color image, where the `clip` function represents the boundary stage function, defined as:

[0128]

[0129] Step S5: Standard image library acquisition and RGB three-channel histogram analysis (corresponding) Figure 1 (RGB histogram analysis of standard image libraries)

[0130] This step provides a target reference for subsequent standardization correction. This embodiment uses a standard reference image set, for example, containing approximately 500 visually appealing fluorescent slide images obtained by a professional physician after adjusting exposure, gain, and gamma parameters. To construct a statistically stable standard reference distribution, this embodiment employs a multi-image histogram aggregation strategy.

[0131] Specifically: Separate each standard image The three channels R, G, B .For example:

[0132]

[0133] After channel separation, calculate the pixel value distribution histogram for each channel (256 bins):

[0134]

[0135] In the above formula, This represents a single-channel histogram. Represents the k-th standard image. Histogram of pixel value distribution in the channel. For indicator functions:

[0136]

[0137] Aggregate the histograms of all N standard images to obtain an aggregate histogram . Normalize the aggregate histogram :

[0138]

[0139] In the above formula, denotes the probability distribution form of the c-th channel aggregate histogram, and the value range . is a dummy variable, i.e. a temporary index in the summation operation, and has no actual physical meaning. The probability distribution .

[0140] Calculate the cumulative distribution function of each of the R, G, and B channels of the standard image:

[0141]

[0142] In the above formula, denotes the cumulative probability distribution of the normalized histogram of the c-th channel. This is the target histogram feature required for the subsequent steps.

[0143] To ensure the numerical stability of the histogram data, define as the effective normalized histogram:

[0144]

[0145] Numerical stability check:

[0146]

[0147] In the above formula, denotes negation, denotes logical OR, is a judgment whether it is NaN, is a judgment whether it is Inf.

[0148] Step S6: RGB three-channel independent histogram specification

[0149] This step performs preliminary color correction on the multi-channel pseudo-color composite image obtained in S4. One of the core features of the present application is that the R, G, and B channels are processed independently.

[0150] The specific steps are as follows: separate the R, G, and B three source channels :

[0151] ​​

[0152] wherein, is the source single channel image, is the source composite image.

[0153] Compute the histogram of each channel of the source image :

[0154]

[0155] Normalize the source channel images to get :

[0156]

[0157] Further compute the cumulative distribution function :

[0158]

[0159] Construct the histogram matching lookup table for each of the R, G, B channels . For each gray value , find the closest corresponding value in the standard CDF:

[0160]

[0161]

[0162] wherein, is the largest integer obtained by binary search that satisfies The monotonicity of the cumulative distribution function (CDF) can be used to quickly locate.

[0163] Apply the LUT transformation to each channel of the source image to get the matched pixel values :

[0164]

[0165] Recombine the three matched channels into the intermediate corrected image :

[0166]

[0167] Then compute the histogram similarity before and after matching:

[0168]

[0169] wherein denotes the matching similarity; The normalized histogram of the matched image. This indicates that the matched histogram is exactly the same as the standard histogram.

[0170] Step S7: Genetic algorithm optimizes the RGB three-channel Gamma parameters (corresponding to...) Figure 1 (Genetic algorithm for optimizing gamma parameter)

[0171] To overcome the shortcomings of traditional Gamma correction parameters that rely on human experience, this invention uses a genetic algorithm (GA) to adaptively search for the optimal combination of Gamma parameters globally.

[0172] Please refer to Figure 3 The diagram illustrates the optimization process of the genetic algorithm in detail. In this preferred embodiment, the GA is designed as follows:

[0173] 1. Individual Encoding and Initialization ( Figure 3 Initial population in ):

[0174] Individual x is encoded as a three-dimensional real vector:

[0175]

[0176] in, This represents the Gamma correction factor for the red channel. This represents the Gamma correction factor for the green channel. This represents the Gamma correction coefficient for the blue channel.

[0177] Set the search space and constraints:

[0178]

[0179] The Gamma value for each channel must be within the specified range. Indicates the lower limit of the interval. Indicates the upper limit of the interval. Typically set to... =0.3, =3.0.

[0180] Randomly generated Individuals (e.g.) ) as the initial population :

[0181]

[0182] Among them, each individual Gamma value Random sampling from a uniform distribution Uniform distribution of intervals Random sampling in the middle, .

[0183] 2. Fitness evaluation (of the fitness evaluation in S6): This is the core of GA. For each individual in the population Figure 3 , first apply Gamma correction operation to the intermediate corrected image obtained in S6 step , for a given Gamma value , to perform a non-linear transformation on the channel :

[0184]

[0185] where is the pixel value before correction, is the pixel value after correction, resulting in the corrected image.

[0186] Then calculate the R, G, B channel histograms of . The fitness function f(x) is used to evaluate the similarity between and the standard histogram obtained in S5.

[0187] In a preferred embodiment, the single-channel comprehensive similarity is a weighted fusion of multiple metrics:

[0188] Pearson correlation coefficient:

[0189]

[0190] where is the similarity coefficient, which measures the degree of overlap between the "standard image channel normalized histogram" and the "Gamma corrected channel normalized histogram". is the standard channel normalized histogram, is the normalized histogram of the corrected image.

[0191] Pearson correlation coefficient ( ) is calculated as follows, which is used to measure the linear correlation between two histograms:

[0192]

[0193] where is defined as the mean of the standard histogram:

[0194]

[0195] Chi-square distance ( ​​The similarity measure of two histograms is as follows, which measures the distribution difference of two histograms:

[0196]

[0197] where is a numerical stability constant.

[0198] The normalized similarity is calculated , which converts the chi-square distance into a similarity:

[0199]

[0200] The three measures are fused by weighting:

[0201]

[0202] where, The core function of is to calculate the sum of the square roots of the products of the corresponding bin values of two normalized histograms, which directly quantifies the distribution similarity between the standard image histogram and the enhanced image histogram, and serves as the main evaluation indicator; The core function of is to measure the linear correlation of two histograms, which can be used for auxiliary evaluation, and Based on the chi-square distance, can be used as a supplement to and Therefore, there is Further, the recommended weights are: , , .

[0203] Consider the weighted contribution of the three channels and the penalty term:

[0204]

[0205] Channel weight (based on human visual perception):

[0206]

[0207] These weights come from standard luminance coefficients.

[0208] The penalty term function :

[0209]

[0210] The penalty term avoids extreme values (too dark or too bright), and the penalty coefficient is used to control the penalty strength.

[0211] 3. Genetic operator​Figure 3 Evolution cycle in Genetic Algorithm:

[0212] Elitism Figure 3 Elitism in Genetic Algorithm: The best individual(s) in each generation is(are) directly passed to the next generation, ensuring the best solution is monotonically non-decreasing. For example:

[0213] Elitism Elitism: The best individual(s) in each generation is(are) directly passed to the next generation, ensuring the best solution is monotonically non-decreasing. For example:

[0214]

[0215] Recommended Elitism ratio: (Keep the best 10% individuals)

[0216] Elitist Set :

[0217]

[0218] In the above equation, is an "individual". In the present invention, an x represents a set of Gamma parameters, and the superscript (i) is just an index of the individual (e.g. the i-th individual); denotes the set of population in the t-th generation; is the fitness value of the i-th individual in the t-th generation. Ranking:

[0219] Ranking:

[0220]

[0221] Elitist Set:

[0222]

[0223] Theoretical guarantee:

[0224] Elitism ensures:

[0225]

[0226] In the above equation, denotes the fitness function; denotes the set of population in the t+1-th generation.

[0227] i.e. the best solution is monotonically non-decreasing.

[0228] Selection Figure 3 Tournament selection in Genetic Algorithm: Tournament selection is used, e.g. randomly select individuals (e.g. 2 individuals) ​​) compose tournament pool, select the highest fitness individual into mating pool.

[0229] Crossover ( Figure 3 single-point crossover in ) : perform single-point crossover with probability ( e.g. = 0.8 ), exchange gene segments ( Gamma values ) of parent individuals.

[0230] Mutation ( Figure 3 Gaussian mutation in ) : perform mutation on each gene position with probability

[0231]

[0232] The above equation represents adding Gaussian noise to each gene value, represents a Gaussian distribution with mean 0 and standard deviation , and boundary handling:

[0233]

[0234] Recommended parameters:

[0235] Mutation probability:

[0236] Mutation standard deviation:

[0237] Adaptive mutation:

[0238] The mutation strength can be dynamically adjusted according to the evolution number:

[0239]

[0240] where is the initial mutation standard deviation ( e.g. ), is the current number of generations, is the decay coefficient.

[0241] 4. Loop and output ( Figure 3 output optimal parameters in ) : repeat the execution of genetic operators for generations ( e.g. generations ), or until the fitness converges ( e.g. less than 1% improvement for 10 consecutive generations ). Finally, the algorithm returns the individual with the highest fitness throughout the entire evolution process as the optimal Gamma correction parameter combination. For example:

[0242] Complete evolution process:

[0243] Initialization: , generate initial population​ ;

[0244]

[0245] 1. Calculate population fitness:

[0246] 2. Record best individual:

[0247] 3. Elitism:

[0248] 4. Tournament selection:

[0249] 5. Crossover operation:

[0250] 6. Mutation operation:

[0251] 7. Build new population:

[0252] 8. Update history:

[0253] end for

[0254] Return:

[0255] Recommended evolution generations:

[0256] Convergence analysis

[0257] Fitness evolution curve:

[0258]

[0259] Convergence criterion:

[0260] (Continuous 10 generations less than 1%)

[0261] Convergence rate:

[0262] Define relative improvement rate:

[0263]

[0264] Fast convergence:

[0265] Slow convergence:

[0266] Convergence:

[0267] Population diversity measure:

[0268]

[0269] where, denotes the Euclidean norm, is the population mean, calculated as:

[0270]

[0271] Global optimum Parameter combination:

[0272]

[0273] Simultaneous output:

[0274] 1. Optimal fitness value: ;

[0275] 2. Fitness evolution history: ;

[0276] 3. Similarity of each channel: (i.e. the optimal individual corresponding value).

[0277] Step S8: RGB three-channel independent Gamma correction and final enhancement (corresponding to RGB three-channel independent gamma correction in Figure 1 )

[0278] This step applies the optimal Gamma value obtained by the genetic algorithm optimization in step S7 to the R, G, B three channels of the intermediate correction image obtained in step S6 for independent Gamma correction.

[0279] Apply the optimal Gamma correction to each channel of the histogram-matched image :

[0280]

[0281] In the above formula, denotes the histogram-matched image, denotes the optimal Gamma parameter.

[0282] And normalize the above formula, let the normalized pixel value be:

[0283]

[0284] Then:

[0285]

[0286] Reduced to [0, 255]:

[0287]

[0288] Next, to improve computational efficiency, pre-compute the Gamma correction lookup table:

[0289]

[0290] In the above formula, represents the pre-computed lookup table for the c-th channel, a vector of length 256.

[0291] Apply Fast transform:

[0292]

[0293] For the pixel values of the matched channel , directly obtain the corrected results through the lookup table .

[0294] Then boundary clipping and type conversion, to ensure that the pixel value is within the valid range:

[0295]

[0296] In the above formula, represents the forced input value limiting in the interval [0, 255], solving the problem of possible pixel value overflow in Gamma correction.

[0297] After type conversion, the final 8-bit unsigned integer image is obtained:

[0298]

[0299] In order to perform color space consistency check, verify the value range distribution of the three channels of RGB:

[0300]

[0301] Statistical verification:

[0302]

[0303]

[0304] In the above formula, represents the minimum value of the image, denotes the maximum value of the image, and verifies the value range of the final image, ensuring that all values fall within the valid range of an 8-bit image [0, 255].

[0305]

[0306]

[0307] In the above formula, denotes the mean value of the channel, reflecting the overall light and dark level of the channel, denotes the standard deviation of the channel, used to reflect the contrast.

[0308] Step S9: Image quality comprehensive evaluation and output

[0309] To verify the effectiveness of the method, the embodiment also includes a quality evaluation step. Evaluation indicators can include peak signal-to-noise ratio (PSNR), structural similarity (SSIM), and comprehensive quality score based on the histogram similarity (HistSim) and the quality score defined in step S7 . The output content includes the final enhanced image , the optimal Gamma parameter ( ), and detailed quality evaluation indicator reports.

[0310]

[0311] wherein the peak signal-to-noise ratio (PSNR) is:

[0312] Mean square error (MSE):

[0313]

[0314] wherein I target (x, y, c) is only a standard reference image (i.e. "gold standard" image) for quality evaluation, and the evaluation step is only for verifying the effectiveness of the method, and is not a necessary component of steps S1-S8.

[0315] PSNR:

[0316]

[0317] in decibels:

[0318]

[0319] Structural similarity (SSIM)

[0320]

[0321] wherein:

[0322] : image mean

[0323] : image standard deviation

[0324] : covariance

[0325] : stability constant

[0326] Decomposition form:

[0327]

[0328] Brightness contrast:

[0329] Contrast contrast:

[0330] Structural contrast:

[0331] Value range: Indicates exactly the same.

[0332] Histogram similarity:

[0333]

[0334] RGB three-channel comprehensive similarity:

[0335]

[0336] Among them , , .

[0337] Comprehensive quality score

[0338]

[0339] Normalized PSNR:

[0340]

[0341] Usually set:

[0342] dB

[0343] dB

[0344] Weight setting:

[0345]

[0346]

[0347] Final output content:

[0348] Enhanced RGB three-channel image: ;

[0349] RGB three independent channel images:

[0350] R channel: ;

[0351] G Channel: ;

[0352] Channel B: ;

[0353] Optimal Gamma parameter: ;

[0354] The evolutionary history of genetic algorithms: ;

[0355] Quality assessment index report: PSNR value, SSIM value, histogram similarity, and overall quality score.

[0356] Figure 4 The presentation demonstrates a comparison of actual enhancement effects using a preferred embodiment of the present invention. The input image was relatively dark, with blurred details and weak fluorescence signals. However, the enhancement results after processing with this method show significant enhancement of cell nuclei (blue) and fluorescent probe signals (red), suppression of background noise, and a substantial improvement in image contrast and clarity, verifying the effectiveness of the present invention.

[0357] Example 2

[0358] Based on the same concept, the present invention also proposes a fluorescent microscopy image intelligent enhancement device for implementing the method of Embodiment 1 above, comprising:

[0359] Image reading module: used to perform image reading, channel detection and separation in step S1.

[0360] Wavelength mapping module: used to perform adaptive wavelength configuration in step S2 and physical pseudo-color mapping in step S3, preferably including LRU cache acceleration function.

[0361] Pseudo-color compositing module: used to perform multi-channel grayscale conversion and maximum value compositing in step S4.

[0362] Standard Analysis Module: Used to execute step S5, extracting RGB three-channel histogram features and CDF from the standard image library.

[0363] Histogram specification module: used to execute step S6, to achieve independent matching and LUT construction of the three RGB channels.

[0364] Genetic optimization module: used to execute step S7, which globally searches for the optimal combination of Gamma parameters through strategies such as multi-objective fitness functions and elite retention.

[0365] Gamma correction module: used to perform step S8, to perform independent correction of the three RGB channels, preferably using LUT acceleration.

[0366] Quality assessment module: used to perform step S9 and calculate multi-dimensional indicators such as PSNR / SSIM.

[0367] Example 3

[0368] This embodiment also provides an electronic device, see reference. Figure 5 It includes a memory 404 and a processor 402, wherein the memory 404 stores a computer program and the processor 402 is configured to run the computer program to perform the steps in any of the above method embodiments.

[0369] Specifically, the processor 402 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement embodiments of the present invention.

[0370] The memory 404 can include a mass storage that stores data or instructions. For example, and without limitation, the memory 404 can include a Hard Disk Drive (HDD), a floppy disk drive, a Solid State Drive (SSD), a flash drive, a Compact Disc Read Only Memory (CD-ROM), a magneto-optical disk, a magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. The memory 404 can be removable and / or non-removable (or fixed) as appropriate. The memory 404 can be internal or external as appropriate. In particular embodiments, the memory 404 is a Non-Volatile memory. In particular embodiments, the memory 404 includes a Read-Only Memory (ROM) and a Random Access Memory (RAM). The ROM can be a mask-programmed ROM, a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically EPROM (EEPROM), an Electrically Alterable ROM (EAROM), or a FLASH memory, or a combination of two or more of these, as appropriate. The RAM can be a Static Random-Access Memory (SRAM) or a Dynamic Random Access Memory (DRAM), which can be a Fast Page Mode Dynamic Random Access Memory (FPMDRAM), an Extended Data Output Dynamic Random Access Memory (EDODRAM), a Synchronous Dynamic Random-Access Memory (SDRAM), or the like, as appropriate.

[0371] The memory 404 can be used to store or buffer various data files needed for processing and / or communication, and possible computer program instructions executed by the processor 402.

[0372] The processor 402 implements any one of the above-mentioned fluorescence microscopic image intelligent enhancement methods by reading and executing the computer program instructions stored in the memory 404.

[0373] Optionally, the above-mentioned electronic device can further include a transmission device 406 connected with the processor 402 and an input / output device 408 connected with the processor 402.

[0374] The transmission device 406 can be used to receive or send data via a network. Specific examples of the network can include a wired or wireless network provided by a communication provider of the electronic device. In one example, the transmission device includes a network adapter (NIC) which can be connected with other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 406 can be a radio frequency (RF) module which is used to communicate with the Internet in a wireless manner.

[0375] The input / output device 408 is used to input or output information.

[0376] Embodiment Four

[0377] The embodiment also provides a readable storage medium, and the readable storage medium stores a computer program. The computer program includes program codes for controlling a process to execute the process. The process includes the fluorescence microscopic image intelligent enhancement method according to Embodiment One.

[0378] It should be noted that specific examples in the embodiment can refer to examples described in the above-mentioned embodiments and optional implementation manners, and the embodiment will not be described here again.

[0379] Generally, various embodiments can be implemented in hardware or special-purpose circuitry, software, logic or any combination thereof. Some aspects of the application can be implemented in hardware, while other aspects can be implemented by firmware or software executed by a controller, microprocessor or other computing device, but the application is not limited thereto. Although various aspects of the application can be illustrated and described as block diagrams, flow charts, or using some other pictorial representation, it is well understood that these blocks, apparatus, systems, techniques or methods described herein can be implemented in hardware, software, firmware, special purpose circuits or logic, general purpose hardware or controller or other computing devices, or some combination thereof.

[0380] Embodiments of the application can be implemented by computer software executable by a data processor of the mobile device such as in the processor entity, or by hardware, or by a combination of software and hardware. Computer software or program, also called program product, including software routines, applets and / or macros, can be stored in any apparatus-readable data storage medium and they include program instructions to implement certain tasks. The program product can include one or more computer-executable components such as by way of example scripts, programs, routines, objects, components, data structures, procedures, subroutines, etc., for implementing one or more embodiments of the application. The computer-executable components can be stored in computer-readable storage medium. The computer-readable storage medium can include one or more of a floppy disk, a ZIP® disk, a hard disk, a video tape, a pen drive, a memory stick, a system memory, a RAM, a ROM, a programmable ROM (EPROM), an erasable programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), a portable computer disk, a Figure 1 Any block in the logical flow of the method described herein, including those of the artificial neural network, can represent a program step, or interconnected logic circuits, blocks and functions, or a combination of program steps and logic circuits, blocks and functions. The software can be stored on such physical media as memory chips, or memory blocks implemented in the processor, magnetic media such as hard disk or floppy disks, and optical media such as for example DVD and the data variants thereof, CD, etc. The physical media are non-transitory media.

[0381] Those skilled in the art should understand that all or some of the technical features in the above embodiments can make any combinations, and in order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present disclosure.

[0382] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as the limitation to the scope of the present application. It should be noted that for those skilled in the art, some modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method of intelligent enhancement of fluorescence microscopy images, characterized in that, The method comprises the following steps: obtaining a multi-channel fluorescence microscopic image and separating the multi-channel fluorescence microscopic image into a plurality of single-channel grayscale images; establishing fluorescence emission wavelengths corresponding to the plurality of single-channel grayscale images respectively, and assigning an emission wavelength to each single-channel grayscale image according to a preset wavelength configuration strategy or a wavelength uniform distribution strategy based on a detected number of channels; based on physical optical characteristics of the fluorescence emission wavelengths, converting the plurality of single-channel grayscale images into corresponding single-channel pseudo-color images respectively, including defining a mapping function of wavelengths covering an ultraviolet to near-infrared waveband to R, G, B three stimulus values, and combining a wavelength-related transparency factor simulating human eye sensitivity to calculate the single-channel pseudo-color images; using a maximum synthesis strategy, taking maximum values of pixel values of the plurality of single-channel pseudo-color images on R, G, B channels respectively to generate a multi-channel pseudo-color synthesis image; obtaining target histogram features of R, G, B three channels of a standard reference image library respectively; respectively performing independent histogram specification processing on R, G, B three channels of the multi-channel pseudo-color synthesis image according to the target histogram features to obtain an intermediate correction image; through an optimization algorithm, taking a similarity degree of an image histogram of the intermediate correction image after Gamma correction and the target histogram features as an optimization target, adaptively searching for and determining optimal Gamma correction parameters corresponding to R, G, B three channels respectively, combining and encoding the Gamma correction parameters of R, G, B three channels into individuals of a genetic algorithm; designing a fitness function, which is used to evaluate a comprehensive similarity degree of R, G, B three channel histograms of the intermediate correction image after the Gamma correction parameters corresponding to the individuals are applied to the intermediate correction image and the target histogram features; through selection, crossover and mutation operations, iteratively evolving a population until convergence or a maximum iteration number is reached to obtain the optimal Gamma correction parameters; applying the optimal Gamma correction parameters to R, G, B three channels respectively, and constructing a lookup table based on the optimal Gamma correction parameters corresponding to R, G, B three channels respectively in advance; using the lookup table to perform pixel value mapping on R, G, B three channels of the intermediate correction image to complete Gamma correction and obtain a final enhanced image; wherein the optimization algorithm is a genetic algorithm; the fitness function comprehensively considers at least one of Bhattacharyya coefficients, Pearson correlation coefficients and chi-square distances to measure the comprehensive similarity degree; the fitness function assigns different weights to similarity degrees of R, G, B three channels when calculating the comprehensive similarity degree, and the weights are determined based on human eye visual perception characteristics.

2. A device for intelligent enhancement of fluorescence microscopy images, characterized in that The method comprises the following steps: a separation module for obtaining a multi-channel fluorescence microscopic image and separating the multi-channel fluorescence microscopic image into a plurality of single-channel grayscale images; a wavelength mapping module for establishing fluorescence emission wavelengths corresponding to the plurality of single-channel grayscale images respectively, and assigning an emission wavelength to each single-channel grayscale image according to a preset wavelength configuration strategy or a wavelength uniform distribution strategy based on a detected number of channels; a pseudo-color conversion module configured to convert the plurality of single-channel grayscale images into corresponding single-channel pseudo-color images based on physical optical characteristics of the fluorescence emission wavelengths, including defining a mapping function of wavelengths covering the ultraviolet to near-infrared wavelength band to R, G, B tristimulus values, and calculating the single-channel pseudo-color images by combining a wavelength-dependent transparency factor simulating human eye sensitivity; an image synthesis module configured to generate a multi-channel pseudo-color composite image by taking maximum values of pixel values of the plurality of single-channel pseudo-color images in R, G, B channels, respectively, using a maximum value synthesis strategy; a standard feature acquisition module configured to acquire target histogram features of R, G, B channels of a standard reference image library; a histogram specification module configured to perform independent histogram specification processing on R, G, B channels of the multi-channel pseudo-color composite image according to the target histogram features, respectively, to obtain intermediate corrected images; a parameter optimization module configured to search and determine optimal Gamma correction parameters of R, G, B channels, respectively, by an optimization algorithm, taking similarity of image histograms of the intermediate corrected images after Gamma correction and the target histogram features as an optimization objective, combining and encoding the Gamma correction parameters of R, G, B channels as individuals of a genetic algorithm; designing a fitness function for evaluating comprehensive similarity of R, G, B channel histograms of the intermediate corrected images after applying the Gamma correction parameters corresponding to the individuals to the intermediate corrected images and the target histogram features; evolving the population by selection, crossover and mutation operations until convergence or reaching a maximum number of iterations to obtain the optimal Gamma correction parameters; a final correction module configured to apply the optimal Gamma correction parameters of R, G, B channels, respectively, to R, G, B channels based on the optimal Gamma correction parameters of R, G, B channels, respectively, to construct look-up tables in advance; mapping pixel values of R, G, B channels of the intermediate corrected images using the look-up tables to complete Gamma correction and obtain a final enhanced image; wherein the optimization algorithm is a genetic algorithm; the fitness function considers at least one of Bhattacharyya coefficient, Pearson correlation coefficient and Chi-square distance to measure the comprehensive similarity; the fitness function assigns different weights to the similarity of R, G, B channels when calculating the comprehensive similarity, and the weights are determined based on human visual perception characteristics. 3.An electronic device comprising a memory and a processor, the electronic device characterized by, The memory stores a computer program, and the processor is configured to run the computer program to execute the fluorescence microscopic image intelligent enhancement method of claim 1.

4. A readable storage medium, characterized by, The readable storage medium stores a computer program, and the computer program includes program code for controlling a process to execute the process, and the process includes the fluorescence microscopic image intelligent enhancement method of claim 1.

Images