A multi-spectral image intelligent fusion method and fusion system

Abstract

The present invention provides a multi-spectral image intelligent fusion method and fusion system, belonging to the field of image processing technology. First, a layered process is performed according to the gray value of a full-color image to obtain multiple gray layers, and the gray variation of each gray layer is calculated to determine its variation level. Based on the variation level, the present invention dynamically adjusts the number of extractions of each gray layer to ensure that important details are retained while reducing unnecessary calculations. Subsequently, by assigning weights to different gray layers and performing weighted fusion, a total pixel set of the full-color image is obtained. Finally, the total pixel set is used to process the multi-spectral image to generate a low-resolution full-color image, and by calculating the ratio of the original full-color image to the low-resolution full-color image, a final full-color and multi-spectral fusion image is generated. The present invention not only improves the resolution and spectral information retention of the fused image, but also reduces the computational complexity.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a multi-spectral image intelligent fusion method and a fusion system. Background Art

[0002] In the field of image processing and remote sensing technology, multispectral image fusion technology is an important technical means, which aims to generate a single image with richer and more detailed information by combining information from different spectral bands. Traditional single-spectral imaging technology, such as using only panchromatic images or multispectral images, often cannot fully reflect the complex characteristics and details of the objects, limiting its application in environmental monitoring, object recognition, medical image analysis and other fields.

[0003] In order to overcome this limitation, researchers began to explore multispectral image fusion technology, combining the high-resolution characteristics of panchromatic images with the multi-band information of multispectral images to achieve a dual improvement in image resolution and spectral information. Panchromatic images are known for their high resolution and can clearly show the texture and details of objects, while multispectral images can capture the characteristics of objects in different spectral bands, such as color, reflectivity, etc., thereby providing richer information about objects.

[0004] However, most of the existing multispectral image fusion methods have problems such as high computational complexity, large information loss or unsatisfactory fusion effect. Traditional fusion methods, such as simple pixel-level superposition or weighted averaging, often fail to effectively balance the resolution and spectral information retention, resulting in poor quality of fused images. In addition, some advanced fusion methods, such as fusion algorithms based on multi-scale transformation or deep learning, although they can achieve good fusion effects, have large computational complexity and high hardware requirements, which limits their popularity in practical applications.

[0005] Therefore, it is necessary to provide a multispectral image intelligent fusion method and fusion system to solve the above technical problems. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides a multispectral image intelligent fusion method and fusion system, which effectively solves the problems of high computational complexity, large information loss and unsatisfactory fusion effect in traditional fusion methods through intelligent layered processing, dynamic adjustment of extraction quantity and weighted fusion strategy.

[0007] The present invention provides a multispectral image intelligent fusion method, which comprises the following steps:

[0008] S1: Acquire a corresponding full-color image and a multispectral image, and perform stratification according to the grayscale value of the full-color image to obtain multiple grayscale layers of the full-color image;

[0009] S2: Calculate the grayscale variation of each grayscale layer, and determine the variation level of the grayscale layer according to a preset variation threshold, wherein the grayscale variation is the standard deviation of the grayscale values ​​of all pixels in the grayscale layer;

[0010] S3: Based on the variation level of each gray layer, dynamically adjust the extraction quantity of each gray layer on the basis of a preset basic ratio, and extract pixels from the corresponding gray layer according to the adjusted extraction quantity to obtain the extracted pixels of the gray layer;

[0011] S4: assigning a weight to each grayscale layer based on the grayscale variation, and weighting the extracted pixels of all grayscale layers based on the weight to obtain a total pixel set of the full-color image;

[0012] S5: Processing the multispectral image based on the total pixel set of the panchromatic image to generate a low-resolution panchromatic image, and calculating the ratio of the original panchromatic image to the low-resolution panchromatic image to generate a panchromatic and multispectral fused image.

[0013] Preferably, step S1 comprises the following steps:

[0014] S101: Acquire a corresponding panchromatic image and a multispectral image to be fused from a data source;

[0015] S102: Acquire a grayscale range of the full-color image, and determine a grayscale span based on the grayscale range;

[0016] S103: Segment the grayscale range of the full-color image based on the grayscale span to obtain a plurality of grayscale layers of the full-color image.

[0017] Preferably, step S2 comprises the following steps:

[0018] S201: traverse each gray level, calculate the gray level mean of the pixels in each gray level, and the mathematical formula of the gray level mean is:

[0019]

[0020] Among them, μ is the grayscale mean, x i is the gray value of the i-th pixel, and n is the number of pixels in the gray layer;

[0021] S202: Based on the grayscale mean, calculate the standard deviation of the grayscale layer, and the mathematical formula of the standard deviation is:

[0022]

[0023] Among them, σ is the standard deviation of the gray level;

[0024] S203: According to a preset variation threshold, the calculated standard deviation is compared with the variation threshold to determine the variation level of the grayscale layer, wherein the variation level includes high variation with a standard deviation greater than the variation threshold, medium variation with a standard deviation equal to the variation threshold, and low variation with a standard deviation less than the variation threshold.

[0025] Preferably, step S3 comprises the following steps:

[0026] S301: dynamically adjusting the extraction quantity of each gray layer according to the variation level of the gray layer and in combination with a preset quantity adjustment rule, wherein the ratio adjustment rule is:

[0027] If the variation level of the grayscale layer is high variation, the number of extractions is increased according to the set increment value based on the preset baseline extraction number.

[0028] If the grayscale variation level is medium variation, the number of baseline extractions remains unchanged.

[0029] If the variation level of the grayscale layer is low variation, then based on the preset baseline extraction number, the extraction number is reduced according to the set decrement value;

[0030] S302: uniformly extract pixels from each grayscale layer using the adjusted extraction quantity to obtain extracted pixels of the grayscale layer.

[0031] Preferably, step S4 comprises the following steps:

[0032] S401: Assign weights to each grayscale layer using a weight assignment formula in combination with the corresponding grayscale variation. The weight assignment formula is:

[0033]

[0034] Among them, ω i is the weight of the i-th grayscale layer, is the pth power of the grayscale variation of the i-th grayscale layer, p is a positive real number used to control the nonlinearity of weight distribution. is the sum of the p-th power of the grayscale variation of all grayscale layers, j is the grayscale layer set, and i∈j;

[0035] S402: traversing the extracted pixels of each grayscale layer and multiplying the grayscale value of each extracted pixel by the weight of the corresponding grayscale layer to obtain a weighted pixel with a weighted grayscale value;

[0036] S403: All weighted pixels of all grayscale layers are integrated to obtain a total pixel set of the full-color image.

[0037] Preferably, step S5 comprises the following steps:

[0038] S501: extracting pixel value information corresponding to the panchromatic image and the upsampled multispectral image from the total pixel set of the panchromatic image, and obtaining fitting coefficients from the multispectral data to the panchromatic data using the extracted pixel value information;

[0039] S502: performing weighted summation on each band of the original multispectral image using the fitting coefficients to generate a low-resolution full-color composite image;

[0040] S503: downsampling the original full-color image to the same resolution as the low-resolution full-color composite image, and calculating the difference between the downsampled full-color image and the low-resolution full-color composite image until the difference does not exceed a preset value;

[0041] S504: Calculate the ratio of the low-resolution full-color composite image and the full-color image corresponding to the difference not exceeding the preset value, and generate a full-color and multi-spectral fused image.

[0042] The present invention also provides a multi-spectral image intelligent fusion system for executing the multi-spectral image intelligent fusion method, the fusion system comprising:

[0043] A grayscale layering module is used to obtain the corresponding full-color image and multispectral image, and to perform layering according to the grayscale value of the full-color image to obtain multiple grayscale layers of the full-color image;

[0044] A variation evaluation module is used to calculate the grayscale variation of each grayscale layer and determine the variation level of the grayscale layer according to a preset variation threshold, wherein the grayscale variation is the standard deviation of the grayscale values ​​of all pixels in the grayscale layer;

[0045] A pixel extraction adjustment module is used to dynamically adjust the extraction quantity of each gray layer based on a preset basic ratio based on the variation level of each gray layer, and extract pixels from the corresponding gray layer according to the adjusted extraction quantity to obtain extracted pixels of the gray layer;

[0046] A weighted fusion module is used to assign a weight to each grayscale layer based on the grayscale variation, and to weight the extracted pixels of all grayscale layers based on the weight to obtain a total pixel set of the full-color image;

[0047] The image fusion generation module is used to generate a low-resolution panchromatic image based on processing the multispectral image from the total pixel set of the panchromatic image, and calculate the ratio of the original panchromatic image and the low-resolution panchromatic image to generate a panchromatic and multispectral fused image.

[0048] Compared with the related art, the multispectral image intelligent fusion method and fusion system provided by the present invention have the following beneficial effects:

[0049] The present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A flowchart of a multispectral image intelligent fusion method provided by the present invention;

[0051] Figure 2 This is a module structure diagram of a multispectral image intelligent fusion system provided by the present invention. DETAILED DESCRIPTION

[0052] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only the parts related to the present invention, rather than all structures, are shown in the accompanying drawings. In addition, the embodiments of the present invention and the features in the embodiments may be combined with each other without conflict.

[0053] It should also be noted that, for ease of description, only the part relevant to the present invention but not all content is shown in the accompanying drawings. It should be mentioned before discussing exemplary embodiments in more detail that some exemplary embodiments are described as processing or methods depicted as flow charts. Although the flow chart describes each operation (or step) as sequential processing, many operations therein can be implemented in parallel, concurrently or simultaneously. In addition, the order of each operation can be rearranged. When its operation is completed, the processing can be terminated, but it can also have additional steps not included in the accompanying drawings. The processing can correspond to methods, functions, procedures, subroutines, subprograms, etc.

[0054] Embodiment 1

[0055] The present invention provides a multi-spectral image intelligent fusion method, referring to Figure 1 As shown, the fusion method comprises the following steps:

[0056] S1: Obtain corresponding full-color images and multi-spectral images, and perform stratification according to the grayscale values ​​of the full-color images to obtain multiple grayscale layers of the full-color images.

[0057] In this embodiment, a data acquisition protocol is used to obtain full-color images and multispectral images of a specified time and place from a database, and the full-color image is converted into a grayscale image, and its grayscale value distribution is read. According to the range of grayscale values, it is evenly divided into multiple grayscale layers, and each layer represents a grayscale value interval, which ensures that the grayscale values ​​of the image are reasonably divided, providing structured grayscale information for subsequent processing.

[0058] S2: Calculate the grayscale variation of each grayscale layer, and determine the variation level of the grayscale layer according to a preset variation threshold, wherein the grayscale variation is the standard deviation of the grayscale values ​​of all pixels in the grayscale layer.

[0059] In this embodiment, for each grayscale layer, the standard deviation of the grayscale values ​​of all pixels within it is calculated to obtain the grayscale variation. The grayscale variation is compared with a preset variation threshold to determine the variation level of each grayscale layer (for example, high variation, medium variation, low variation). The calculation of grayscale variation helps to identify complex and simple areas in the image and provide guidance for subsequent pixel extraction.

[0060] S3: Based on the variation level of each gray layer, dynamically adjust the extraction quantity of each gray layer on the basis of a preset basic ratio, and extract pixels from the corresponding gray layer according to the adjusted extraction quantity to obtain the extracted pixels of the gray layer.

[0061] In this embodiment, a basic pixel extraction ratio is preset. For example, 10% of the pixels are extracted from each gray level.

[0062] The ratio is adjusted according to the variation level. The extraction ratio is increased for grayscale layers with high variation, and reduced for grayscale layers with low variation. A corresponding number of pixels are extracted from each grayscale layer to form an extracted pixel set. This ensures that the information in complex areas is more fully retained, and simplified areas reduce redundancy and improve fusion efficiency.

[0063] S4: Assigning a weight to each grayscale layer based on the grayscale variation, and weighting the extracted pixels of all grayscale layers based on the weight to obtain a total pixel set of the full-color image.

[0064] In this embodiment, the grayscale variation is converted into weights. The grayscale layer with high variation has a higher weight, and vice versa. The extracted pixels of all grayscale layers are weighted according to the weights to form the total pixel set of the full-color image. The weighted process enables the details of the image to be represented more accurately, thereby improving the quality of the fused image.

[0065] S5: Processing the multispectral image based on the total pixel set of the panchromatic image to generate a low-resolution panchromatic image, and calculating the ratio of the original panchromatic image to the low-resolution panchromatic image to generate a panchromatic and multispectral fused image.

[0066] In this embodiment, a low-resolution panchromatic image is reconstructed using the total pixel set, and the ratio between the original panchromatic image and the low-resolution panchromatic image, usually called the brightness scaling factor, is calculated. This scaling factor is applied to the multispectral image to adjust its brightness, and then it is fused with the panchromatic image. The fusion process takes into account the consistency of brightness, and the generated fused image has true colors and rich details, and is suitable for a variety of remote sensing applications.

[0067] Specifically, step S1 includes the following steps:

[0068] S101: Obtain corresponding panchromatic images and multispectral images to be fused from a data source.

[0069] In this embodiment, the data source is connected: FTP, HTTP, or HTTPS communication protocols are used to communicate with the data source device.

[0070] Request Data: Send a request to obtain image data for a specific location, date, and time, ensuring that the timestamps of the panchromatic and multispectral images are as close as possible to reduce errors caused by time differences.

[0071] Receiving and storing images: After receiving the image data, store it on a local hard drive or in a cloud storage service for further processing.

[0072] S102: Acquire the grayscale range of the full-color image, and determine the grayscale span based on the grayscale range.

[0073] In this embodiment, full-color image data is read: the data matrix of the full-color image is read using an image processing library.

[0074] Calculate the grayscale range: traverse the data matrix and find the minimum and maximum grayscale values, which constitute the two ends of the grayscale range.

[0075] Determine the grayscale span: select an appropriate number of intervals (exemplarily, it can be 16, 32 or 64), and then calculate the grayscale span, that is, the width of each grayscale interval, based on the minimum and maximum values ​​of the grayscale range.

[0076] For example, if the grayscale range is from 0 to 255, and the number of intervals is set to 16, then each grayscale span will be 16 (ie, 256 / 16=16).

[0077] S103: Segment the grayscale range of the full-color image based on the grayscale span to obtain a plurality of grayscale layers of the full-color image.

[0078] In this embodiment, an empty list is created to store the segmented grayscale layers.

[0079] Segmentation of grayscale layers: According to the determined grayscale span, the grayscale value range of the full-color image is divided into multiple continuous intervals, and each interval constitutes a grayscale layer.

[0080] Extract grayscale layer: For each grayscale interval, use threshold processing technology to extract the pixels in the interval to form a new matrix containing only the pixels in the interval, and set the remaining pixels to 0 or ignore them to create a grayscale layer.

[0081] Store grayscale layers: Add the matrix of each grayscale layer to the grayscale layer list for use in subsequent steps.

[0082] Specifically, step S2 includes the following steps:

[0083] S201: traverse each gray level, calculate the gray level mean of the pixels in each gray level, and the mathematical formula of the gray level mean is:

[0084]

[0085] Among them, μ is the grayscale mean, x i is the gray value of the i-th pixel, and n is the number of pixels in the gray layer.

[0086] In this embodiment, a variable is initialized to save the grayscale mean, which is usually set to 0. Each pixel in the grayscale layer is traversed and its grayscale value is accumulated. After the traversal is completed, the total accumulated grayscale value is divided by the number of pixels to obtain the grayscale mean. The grayscale mean is calculated for the subsequent calculation of the standard deviation. The grayscale mean reflects the overall brightness level of the grayscale layer and provides a benchmark for evaluating the variability.

[0087] S202: Based on the grayscale mean, calculate the standard deviation of the grayscale layer, and the mathematical formula of the standard deviation is:

[0088]

[0089] Among them, σ is the standard deviation of the grayscale layer.

[0090] In this embodiment, a variable is initialized again to save the sum of square differences, which is usually set to 0. The grayscale mean calculated previously is used to traverse each pixel in the grayscale layer, and the square of the difference between the grayscale value of each pixel and the grayscale mean is calculated. These square differences are then accumulated. After the traversal is completed, the accumulated sum of square differences is divided by the number of pixels, and the square root is taken to obtain the standard deviation. The standard deviation measures the degree of discreteness of the grayscale values ​​of the pixels in the grayscale layer. The larger the standard deviation, the more complex the grayscale changes in the grayscale layer. Conversely, the changes are smaller and more uniform.

[0091] S203: According to a preset variation threshold, the calculated standard deviation is compared with the variation threshold to determine the variation level of the grayscale layer, wherein the variation level includes high variation with a standard deviation greater than the variation threshold, medium variation with a standard deviation equal to the variation threshold, and low variation with a standard deviation less than the variation threshold.

[0092] In this embodiment, a variation threshold is set, which can be determined based on experience or experiments to distinguish different variation levels.

[0093] Compare the standard deviation and variation threshold calculated for each grayscale layer.

[0094] If σ is greater than the variation threshold, the grayscale layer belongs to the high variation level.

[0095] If σ is equal to the variation threshold, the grayscale layer belongs to the medium variation level.

[0096] If σ is less than the variation threshold, the grayscale layer belongs to the low variation level.

[0097] Mark or record the variation level of each grayscale layer for use in subsequent steps. The determination of the variation level helps to intelligently adjust the pixel extraction strategy in subsequent steps to ensure that more details of complex areas (high variation) are retained, while simple areas (low variation) reduce redundancy, thereby improving the quality and efficiency of image fusion.

[0098] Specifically, step S3 includes the following steps:

[0099] S301: dynamically adjusting the extraction quantity of each gray layer according to the variation level of the gray layer and in combination with a preset quantity adjustment rule, wherein the ratio adjustment rule is:

[0100] If the variation level of the grayscale layer is high variation, the number of extractions is increased according to the set increment value based on the preset baseline extraction number.

[0101] If the grayscale variation level is medium variation, the number of baseline extractions remains unchanged.

[0102] If the variation level of the grayscale layer is low variation, the number of extractions is reduced according to the set decrement value based on the preset baseline extraction number.

[0103] In this embodiment, a baseline extraction number is set, for example, assuming that 10% of the pixels are initially extracted from each gray level.

[0104] The increment value and the decrement value are set, the increment value is used to increase the number of extractions of the grayscale layer with high variability, and the decrement value is used to reduce the number of extractions of the grayscale layer with low variability. Exemplarily, the increment value and the decrement value can be 10%.

[0105] For each grayscale layer:

[0106] If the grayscale layer has a high degree of variation, increase the number of pixels in the incremental value based on the number of baseline extractions.

[0107] If the grayscale variation level is medium, keep the number of baseline extractions unchanged.

[0108] If the grayscale layer has a low degree of variation, the number of pixels with reduced values ​​is reduced based on the number of baseline extractions.

[0109] Update the number of decimations for each grayscale layer.

[0110] By dynamically adjusting the number of extractions, we can ensure that complex areas (high variability) in the image receive more attention, thereby better preserving details during the fusion process; at the same time, we can reduce redundant information in simple areas (low variability), optimize the use of computing resources, and improve processing efficiency.

[0111] S302: uniformly extract pixels from each grayscale layer using the adjusted extraction quantity to obtain extracted pixels of the grayscale layer.

[0112] In this embodiment, for each gray level, pixel extraction is performed according to the adjusted extraction quantity.

[0113] Create a random number generator that selects cells uniformly at random.

[0114] Pixels are randomly selected from the grayscale layer until the adjusted number of decimations is reached.

[0115] The extracted pixels are stored in a set, which is called the extracted pixels of the grayscale layer.

[0116] Uniform pixel extraction ensures the representativeness of each grayscale layer and avoids any potential bias or concentrated extraction of certain specific areas. This method can fairly extract information from each grayscale layer, which is beneficial to the subsequent fusion processing and ensures the accuracy and integrity of the fused image.

[0117] Specifically, step S4 includes the following steps:

[0118] S401: Assign weights to each grayscale layer using a weight assignment formula in combination with the corresponding grayscale variation. The weight assignment formula is:

[0119]

[0120] Among them, ω i is the weight of the i-th grayscale layer, is the pth power of the grayscale variation of the i-th grayscale layer, p is a positive real number used to control the nonlinearity of weight distribution. is the sum of the p-th power of the grayscale variation of all grayscale layers, j is the grayscale layer set, and i∈j.

[0121] In this embodiment, a weight is assigned to each grayscale layer by comparing the pth power of the grayscale variability with the sum of the pth powers of the variability of all grayscale layers, so that grayscale layers with high variability will obtain higher weights, while grayscale layers with low variability have relatively smaller weights.

[0122] S402: By traversing the extracted pixels of each grayscale layer, and multiplying the grayscale value of each extracted pixel by the weight of the corresponding grayscale layer, a weighted pixel with a weighted grayscale value is obtained.

[0123] In this embodiment, for the extracted pixels of each gray level, the following operations are performed:

[0124] Get the grayscale value of the extracted pixel.

[0125] Use the weight of the corresponding grayscale layer calculated in step S401.

[0126] Multiply the grayscale value by the weight to get the weighted grayscale value.

[0127] This weighted gray value is the value of the weighted pixel.

[0128] S403: All weighted pixels of all grayscale layers are integrated to obtain a total pixel set of the full-color image.

[0129] In this embodiment, after completing the weighted pixel calculations of all gray levels, the following operations are performed:

[0130] Traverse all weighted pixels on all grayscale layers to construct the total pixel set of the full-color image.

[0131] Specifically, step S5 includes the following steps:

[0132] S501: extracting pixel value information corresponding to the panchromatic image and the upsampled multispectral image from the total pixel set of the panchromatic image, and obtaining fitting coefficients from the multispectral data to the panchromatic data using the extracted pixel value information.

[0133] In this embodiment, pixel values ​​at the same position of the panchromatic image and the upsampled multispectral image are extracted.

[0134] Using these pixel values, a linear model is found through the least squares method. This model can describe how multispectral data is mapped to panchromatic data. The coefficients of this model are the fitting coefficients, which will be used for the subsequent weighted summation of multispectral images.

[0135] S502: Performing weighted summation on each band of the original multispectral image using the fitting coefficients to generate a low-resolution full-color composite image.

[0136] In this embodiment, after the fitting coefficients are obtained, a low-resolution full-color composite image can be generated by weighted summing each band of the original multispectral image. The resolution of this composite image is low because its goal is to match the resolution of the downsampled full-color image in the subsequent step.

[0137] S503: down-sampling the original full-color image to the same resolution as the low-resolution full-color composite image, and calculating the difference between the down-sampled full-color image and the low-resolution full-color composite image until the difference does not exceed a preset value.

[0138] In this embodiment, in order to adjust the difference between the full-color image and the multi-spectral composite image to make them as consistent as possible, the full-color image is downsampled to match the resolution of the two, and then compared and adjusted until the difference reaches an acceptable standard. This preset value is a threshold defined in advance, indicating the maximum difference allowed.

[0139] S504: Calculate the ratio of the low-resolution full-color composite image and the full-color image corresponding to the difference not exceeding the preset value, and generate a full-color and multi-spectral fused image.

[0140] In this embodiment, once the difference between the low-resolution panchromatic composite image and the downsampled panchromatic image meets the requirement, the ratio between the two can be calculated. This ratio can be used to guide how to fuse the high-resolution panchromatic image and the multispectral image to generate the final panchromatic and multispectral fused image.

[0141] The whole process is an iterative optimization process, through continuous adjustment and comparison, until the best fusion effect is achieved. The final fused image will contain the spatial details of the panchromatic image and the spectral information of the multispectral image, forming an image with both high resolution and rich color levels.

[0142] The working principle of a multispectral image intelligent fusion method provided by the present invention is as follows: an efficient fusion of a panchromatic image and a multispectral image is achieved through an intelligent fusion strategy. The method first performs layered processing according to the grayscale value of the panchromatic image to obtain multiple grayscale layers, and calculates the grayscale variation of each grayscale layer to determine its variation level. Based on the variation level, the present invention dynamically adjusts the number of extractions of each grayscale layer to ensure that unnecessary calculations are reduced while retaining important details. Subsequently, by assigning weights to different grayscale layers and performing weighted fusion, the total pixel set of the panchromatic image is obtained. Finally, the multispectral image is processed using the total pixel set to generate a low-resolution panchromatic image, and the final panchromatic and multispectral fusion image is generated by calculating the ratio of the original panchromatic image to the low-resolution panchromatic image.

[0143] The multispectral image intelligent fusion method of the present invention effectively solves the problems of high computational complexity, large information loss and unsatisfactory fusion effect in traditional fusion methods through intelligent layered processing, dynamic adjustment of extraction quantity and weighted fusion strategy. This method not only improves the resolution and spectral information retention of the fused image, but also reduces the computational complexity, providing a more efficient and accurate image fusion solution for remote sensing, medical image analysis and other fields.

[0144] Embodiment 2

[0145] The present invention also provides a multi-spectral image intelligent fusion system for executing the multi-spectral image intelligent fusion method. Figure 2 As shown, the fusion system comprises:

[0146] The grayscale layering module 100 is used to obtain the corresponding full-color image and multi-spectral image, and perform layering according to the grayscale value of the full-color image to obtain multiple grayscale layers of the full-color image.

[0147] The variation evaluation module 200 is used to calculate the grayscale variation of each grayscale layer and determine the variation level of the grayscale layer according to a preset variation threshold, wherein the grayscale variation is the standard deviation of the grayscale values ​​of all pixels in the grayscale layer.

[0148] The pixel extraction adjustment module 300 is used to dynamically adjust the extraction quantity of each gray layer based on the variation level of each gray layer and on the basis of a preset basic ratio, and extract pixels from the corresponding gray layer according to the adjusted extraction quantity to obtain the extracted pixels of the gray layer.

[0149] The weighted fusion module 400 is used to assign a weight to each grayscale layer based on the grayscale variation, and weight the extracted pixels of all grayscale layers based on the weight to obtain a total pixel set of the full-color image.

[0150] The image fusion generation module 500 is used to generate a low-resolution panchromatic image based on processing the multispectral image from the total pixel set of the panchromatic image, and calculate the ratio of the original panchromatic image and the low-resolution panchromatic image to generate a panchromatic and multispectral fused image.

[0151] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0152] A person skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0153] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. A multispectral image intelligent fusion method, characterized in that: The fusion method comprises the following steps: S1: Acquire a corresponding full-color image and a multispectral image, and perform stratification according to the grayscale value of the full-color image to obtain multiple grayscale layers of the full-color image; S2: Calculate the grayscale variation of each grayscale layer, and determine the variation level of the grayscale layer according to a preset variation threshold, wherein the grayscale variation is the standard deviation of the grayscale values ​​of all pixels in the grayscale layer; S3: Based on the variation level of each gray layer, dynamically adjust the extraction quantity of each gray layer on the basis of a preset basic ratio, and extract pixels from the corresponding gray layer according to the adjusted extraction quantity to obtain the extracted pixels of the gray layer; S4: assigning a weight to each grayscale layer based on the grayscale variation, and weighting the extracted pixels of all grayscale layers based on the weight to obtain a total pixel set of the full-color image; S5: Processing the multispectral image based on the total pixel set of the panchromatic image to generate a low-resolution panchromatic image, and calculating the ratio of the original panchromatic image to the low-resolution panchromatic image to generate a panchromatic and multispectral fused image.

2. The multispectral image intelligent fusion method according to claim 1, characterized in that: Step S1 includes the following steps: S101: Acquire a corresponding panchromatic image and a multispectral image to be fused from a data source; S102: Acquire a grayscale range of the full-color image, and determine a grayscale span based on the grayscale range; S103: Segment the grayscale range of the full-color image based on the grayscale span to obtain a plurality of grayscale layers of the full-color image.

3. The multispectral image intelligent fusion method according to claim 2, characterized in that: Step S2 includes the following steps: S201: traverse each gray level, calculate the gray level mean of the pixels in each gray level, and the mathematical formula of the gray level mean is: Among them, μ is the grayscale mean, x i is the gray value of the i-th pixel, and n is the number of pixels in the gray layer; S202: Based on the grayscale mean, calculate the standard deviation of the grayscale layer, and the mathematical formula of the standard deviation is: Among them, σ is the standard deviation of the gray level; S203: According to a preset variation threshold, the calculated standard deviation is compared with the variation threshold to determine the variation level of the grayscale layer, wherein the variation level includes high variation with a standard deviation greater than the variation threshold, medium variation with a standard deviation equal to the variation threshold, and low variation with a standard deviation less than the variation threshold.

4. The multispectral image intelligent fusion method according to claim 3, characterized in that: Step S3 includes the following steps: S301: dynamically adjusting the extraction quantity of each gray layer according to the variation level of the gray layer and in combination with a preset quantity adjustment rule, wherein the ratio adjustment rule is: If the variation level of the grayscale layer is high variation, the number of extractions is increased according to the set increment value based on the preset baseline extraction number. If the grayscale variation level is medium variation, the number of baseline extractions remains unchanged. If the variation level of the grayscale layer is low variation, then based on the preset baseline extraction number, the extraction number is reduced according to the set decrement value; S302: uniformly extract pixels from each grayscale layer using the adjusted extraction quantity to obtain extracted pixels of the grayscale layer.

5. The multispectral image intelligent fusion method according to claim 4, characterized in that: Step S4 includes the following steps: S401: Assign weights to each grayscale layer using a weight assignment formula in combination with the corresponding grayscale variation. The weight assignment formula is: Among them, ω i is the weight of the i-th grayscale layer, is the pth power of the grayscale variation of the i-th grayscale layer, p is a positive real number used to control the nonlinearity of weight distribution. is the sum of the i-th power of the grayscale variation of all grayscale layers, j is the grayscale layer set, and i∈j; S402: traversing the extracted pixels of each grayscale layer and multiplying the grayscale value of each extracted pixel by the weight of the corresponding grayscale layer to obtain a weighted pixel with a weighted grayscale value; S403: All weighted pixels of all grayscale layers are integrated to obtain a total pixel set of the full-color image.

6. The multispectral image intelligent fusion method according to claim 5, characterized in that: Step S5 includes the following steps: S501: extracting pixel value information corresponding to the panchromatic image and the upsampled multispectral image from the total pixel set of the panchromatic image, and obtaining fitting coefficients from the multispectral data to the panchromatic data using the extracted pixel value information; S502: performing weighted summation on each band of the original multispectral image using the fitting coefficients to generate a low-resolution full-color composite image; S503: downsampling the original full-color image to the same resolution as the low-resolution full-color composite image, and calculating the difference between the downsampled full-color image and the low-resolution full-color composite image until the difference does not exceed a preset value; S504: Calculate the ratio of the low-resolution full-color composite image and the full-color image corresponding to the difference not exceeding the preset value, and generate a full-color and multi-spectral fused image.

7. A multispectral image intelligent fusion system, used to execute a multispectral image intelligent fusion method according to any one of claims 1 to 6, characterized in that: The fusion system comprises: A grayscale layering module is used to obtain the corresponding full-color image and multispectral image, and to perform layering according to the grayscale value of the full-color image to obtain multiple grayscale layers of the full-color image; A variation evaluation module is used to calculate the grayscale variation of each grayscale layer and determine the variation level of the grayscale layer according to a preset variation threshold, wherein the grayscale variation is the standard deviation of the grayscale values ​​of all pixels in the grayscale layer; A pixel extraction adjustment module is used to dynamically adjust the extraction quantity of each gray layer based on a preset basic ratio based on the variation level of each gray layer, and extract pixels from the corresponding gray layer according to the adjusted extraction quantity to obtain extracted pixels of the gray layer; A weighted fusion module is used to assign a weight to each grayscale layer based on the grayscale variation, and to weight the extracted pixels of all grayscale layers based on the weight to obtain a total pixel set of the full-color image; The image fusion generation module is used to generate a low-resolution panchromatic image based on processing the multispectral image from the total pixel set of the panchromatic image, and calculate the ratio of the original panchromatic image and the low-resolution panchromatic image to generate a panchromatic and multispectral fused image.

Images