A multi-color visual image enhancement method and system based on image fusion

Abstract

The present application provides a multi-color visual image enhancement method and system based on image fusion, which relates to the field of image enhancement technology. All multi-color visual images of paper to be detected are obtained; the selected multi-color visual image is subjected to step-by-step noise suppression through a preset sliding window to obtain a noise-suppressed image, the horizontal feature difference amount and the vertical feature difference amount are determined, and the noise-suppressed image is converted into a multi-color visual sharpened image; the detail information fusion factor is determined from the multi-color visual sharpened image, and descriptor extraction is performed on the multi-color visual sharpened image to obtain a visual feature descriptor; the corresponding visual feature descriptors are subjected to feature fusion based on the corresponding detail information fusion factor to obtain a feature-fused visual feature descriptor, and a multi-color visual enhanced image of the paper to be detected is generated based on the feature-fused visual feature descriptor. The present application can improve the enhancement effect of the multi-color visual image in the fusion process to enhance the credibility of paper quality detection.

Description

Technical Field

[0001] The present application relates to the technical field of image enhancement, and more specifically, to a multi-color visual image enhancement method and system based on image fusion. Background Art

[0002] Multicolor visual image enhancement based on image fusion has important application value in the papermaking and paper products industry. This technology improves image quality by fusing multicolor information from different imaging sources, thereby more accurately detecting the quality of paper and its products, optimizing printing effects, and enhancing anti-counterfeiting characteristics. It uses multi-dimensional image data such as different color channels, texture features, and edge information to integrate into an image containing richer visual information, thereby improving the brightness, contrast, clarity, and detail performance of the image, thereby achieving the purpose of enhancing visual effects.

[0003] In the process of papermaking and paper product inspection, images are often disturbed by noise to varying degrees, such as Gaussian noise introduced by dust and illumination changes in the production line environment, salt and pepper noise caused by fiber inhomogeneity on the paper surface, and quantization noise that may be generated during scanning or shooting. These noises not only affect the detection accuracy of paper surface defects (such as bubbles, holes, creases, and impurities), but also easily introduce unnecessary pseudo-information during the image fusion process, resulting in a decrease in the quality of the final enhancement result. For example, in the analysis of paper fiber distribution, noise may mask fine texture structures and affect the reliability of the automatic detection system. In addition, the selection of fusion weights is also a key factor affecting the fusion effect. In different application scenarios in the papermaking industry, different image features have different importance. For example, in the detection of paper thickness uniformity, infrared imaging information may be more important than visible light, while in the analysis of printed color consistency, the information weight of the RGB channel should be higher. Therefore, how to improve the enhancement effect of multi-color visual images in the fusion process to enhance the credibility of paper quality detection is a key technical challenge to improve the quality control capabilities of papermaking and paper products. Summary of the invention

[0004] The present application provides a multi-color visual image enhancement method and system based on image fusion, which can improve the enhancement effect of the multi-color visual image during the fusion process to enhance the credibility of paper quality detection.

[0005] In a first aspect, the present application provides a multi-color visual image enhancement method based on image fusion, the image enhancement method comprising the following steps:

[0006] Acquire all multi-color visual images of the paper to be detected, and select one multi-color visual image from all the multi-color visual images;

[0007] Performing step-by-step noise suppression on the selected multi-color visual image through a preset sliding window to obtain a noise suppressed image corresponding to the selected multi-color visual image, and then determining a horizontal feature difference amount and a vertical feature difference amount of the noise suppressed image, and converting the noise suppressed image into a multi-color visual sharpened image according to the horizontal feature difference amount and the vertical feature difference amount;

[0008] Determining a detail information fusion factor corresponding to a selected multi-color visual image from the multi-color visually sharpened image, performing descriptor extraction on the multi-color visually sharpened image, and then obtaining a visual feature descriptor corresponding to the selected multi-color visual image;

[0009] Continue to select other multi-color visual images, and then obtain the detail information fusion factor and the corresponding visual feature descriptor corresponding to each multi-color visual image, based on the corresponding detail information fusion factor, perform feature fusion on the visual feature descriptor corresponding to each multi-color visual image to obtain the feature-fused visual feature descriptor, and generate a multi-color visual enhanced image of the paper to be detected based on the feature-fused visual feature descriptor.

[0010] In this embodiment, all multi-color visual images of the paper to be detected are acquired by an image sensor.

[0011] In this embodiment, performing step-by-step noise suppression on the selected multi-color visual image through a preset sliding window to obtain a noise suppressed image corresponding to the selected multi-color visual image specifically includes:

[0012] Suppressing the correlated noise of the selected multi-color visual image according to the sliding window, thereby obtaining a multi-color visual image after suppressing the correlated noise;

[0013] The uncorrelated noise is suppressed on the polychromatic visual image after the correlated noise is suppressed, so as to obtain a noise suppressed image corresponding to the selected polychromatic visual image.

[0014] In this embodiment, the horizontal feature difference and the vertical feature difference of the noise suppressed image are determined by taking the pixel standard deviation of the noise suppressed image in the horizontal direction as the horizontal feature difference of the noise suppressed image, and taking the pixel standard deviation of the noise suppressed image in the vertical direction as the vertical feature difference of the noise suppressed image.

[0015] In this embodiment, converting the noise suppressed image into a multi-color visually sharpened image according to the horizontal feature difference amount and the vertical feature difference amount specifically includes:

[0016] Determine the aspect ratio by using the horizontal feature difference and the vertical feature difference;

[0017] The noise suppression image is convolved according to the aspect difference ratio, the horizontal feature difference amount and the vertical feature difference amount, so as to obtain a multi-color visually sharpened image.

[0018] In this embodiment, determining the detail information fusion factor corresponding to the selected multi-color visual image from the multi-color visual sharpened image specifically includes:

[0019] Get the preset local window;

[0020] determining all local information entropies in the polychromatic visually sharpened image based on the local window;

[0021] The detail information fusion factor corresponding to the selected multi-color visual image is determined according to all local information entropies.

[0022] In this embodiment, extracting descriptors from the multi-color visually sharpened image is performed by extracting descriptors from the multi-color visually sharpened image through a recurrent neural network.

[0023] In this embodiment, the visual feature descriptors corresponding to each multi-color visual image are feature fused based on the corresponding detail information fusion factor, and the visual feature descriptors obtained after feature fusion specifically include:

[0024] The detail information fusion factor corresponding to each multi-color visual image and the corresponding visual feature descriptor are multiplied respectively, and the sum of all multiplication results is used as the visual feature descriptor after feature fusion.

[0025] In this embodiment, the multi-color visual enhanced image of the paper to be detected is generated based on the visual feature descriptor after the feature fusion, and the visual feature descriptor after the feature fusion is mapped to the image space using the reverse mapping technology to obtain the multi-color visual enhanced image of the paper to be detected.

[0026] In a second aspect, the present application provides a multi-color visual image enhancement system based on image fusion, which is used to perform a multi-color visual image enhancement method based on image fusion, and the image enhancement system includes:

[0027] An image acquisition module is used to acquire all multi-color visual images of the paper to be detected and select one multi-color visual image from all the multi-color visual images;

[0028] A noise suppression module is used to perform step-by-step noise suppression on the selected multi-color visual image through a preset sliding window to obtain a noise suppressed image corresponding to the selected multi-color visual image, and then determine a horizontal feature difference amount and a vertical feature difference amount of the noise suppressed image, and convert the noise suppressed image into a multi-color visual sharpened image according to the horizontal feature difference amount and the vertical feature difference amount;

[0029] A feature extraction module, used to determine a detail information fusion factor corresponding to a selected multi-color visual image from the multi-color visually sharpened image, extract a descriptor from the multi-color visually sharpened image, and then obtain a visual feature descriptor corresponding to the selected multi-color visual image;

[0030] The image enhancement module is used to continue selecting other multi-color visual images, and then obtain the detail information fusion factor and the corresponding visual feature descriptor corresponding to each multi-color visual image, perform feature fusion on the visual feature descriptor corresponding to each multi-color visual image based on the corresponding detail information fusion factor, obtain the feature-fused visual feature descriptor, and generate a multi-color visual enhanced image of the paper to be detected based on the feature-fused visual feature descriptor.

[0031] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects:

[0032] By acquiring all the multi-color visual images of the paper to be detected, a multi-color visual image is selected from all the multi-color visual images; noise suppression is performed on the selected multi-color visual image in steps through a preset sliding window to obtain a noise suppression image corresponding to the selected multi-color visual image, and then the horizontal feature difference amount and the vertical feature difference amount of the noise suppression image are determined, and the noise suppression image is converted into a multi-color visual sharpened image according to the horizontal feature difference amount and the vertical feature difference amount; the detail information fusion factor corresponding to the selected multi-color visual image is determined from the multi-color visual sharpened image, and descriptor extraction is performed on the multi-color visual sharpened image to obtain a visual feature descriptor corresponding to the selected multi-color visual image; other multi-color visual images are continuously selected to obtain a detail information fusion factor and a corresponding visual feature descriptor corresponding to each multi-color visual image, and the visual feature descriptor corresponding to each multi-color visual image is feature fused based on the corresponding detail information fusion factor to obtain a visual feature descriptor after feature fusion, and a multi-color visual enhanced image of the paper to be detected is generated according to the visual feature descriptor after feature fusion.

[0033] It can be seen that in the present application, denoising can be effectively performed during the fusion process, and a suitable fusion ratio can be adaptively selected; wherein, through the step-by-step noise suppression of the sliding window, noise can be removed in a targeted manner in the local area, and after noise suppression, the image details and texture changes can be quantified by calculating the pixel standard deviation of the image in the horizontal and vertical directions (i.e., the horizontal feature difference and the vertical feature difference), and the degree of sharpening processing can be adjusted by the horizontal feature difference and the vertical feature difference, which helps to avoid excessive sharpening of the image; then, the detail information fusion factor can be used as a weighting coefficient to determine the contribution ratio of different regions in the fusion, and the visual feature descriptor of the image is extracted by a recurrent neural network, which can effectively extract high-dimensional and discriminative features from the image; finally, a multi-color visual enhanced image is generated by a feature fusion method based on the detail information fusion factor, which can selectively fuse the detail information in different multi-color visual images into the final image to avoid image distortion due to excessive enhancement.

[0034] In summary, the technical solution adopted in this application can improve the enhancement effect of multi-color visual images during the fusion process to enhance the credibility of paper quality detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0036] Figure 1 is a flow chart of a multi-color visual image enhancement method based on image fusion provided by the present application;

[0037] Figure 2 is an exemplary flow chart for determining a multi-color visually sharpened image according to the present application;

[0038] Figure 3 is an exemplary flow chart for determining a detail information fusion factor corresponding to a selected multi-color visual image provided by the present application;

[0039] Figure 4 It is a module structure diagram of a multi-color visual image enhancement system based on image fusion provided by the present application. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0041] The embodiment of the present application provides a multi-color visual image enhancement method and system based on image fusion, the core of which is to obtain all multi-color visual images of a paper to be detected, select a multi-color visual image from all the multi-color visual images; perform step-by-step noise suppression on the selected multi-color visual image through a preset sliding window to obtain a noise suppression image corresponding to the selected multi-color visual image, and then determine the horizontal feature difference amount and the vertical feature difference amount of the noise suppression image, and convert the noise suppression image into a multi-color visual sharpened image according to the horizontal feature difference amount and the vertical feature difference amount; determine the detail information fusion factor corresponding to the selected multi-color visual image from the multi-color visual sharpened image, perform descriptor extraction on the multi-color visual sharpened image, and then obtain the visual feature descriptor corresponding to the selected multi-color visual image; continue to select other multi-color visual images, and then obtain the detail information fusion factor and the corresponding visual feature descriptor corresponding to each multi-color visual image, perform feature fusion on the visual feature descriptor corresponding to each multi-color visual image based on the corresponding detail information fusion factor, obtain the visual feature descriptor after feature fusion, and generate a multi-color visual enhanced image of the paper to be detected according to the visual feature descriptor after feature fusion. The above scheme can improve the enhancement effect of multi-color visual images during the fusion process to enhance the credibility of paper quality detection.

[0042] Embodiment 1: In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods. Figure 1 As shown in FIG. 1 , this figure is an exemplary flow chart of a multi-color visual image enhancement method based on image fusion according to this embodiment of the present application, and the image enhancement method includes the following steps:

[0043] In step S1, all multi-color visual images of the paper to be detected are acquired, and one multi-color visual image is selected from all the multi-color visual images.

[0044] In specific implementation, all multi-color visual images of the paper to be detected can be obtained through an image sensor; the image sensor used in this application is a charge-coupled device sensor; after all the multi-color visual images are obtained, one multi-color visual image is selected as the processing and description object, and other multi-color visual images can also be processed in the same way.

[0045] In step S2, the selected multi-color visual image is subjected to step-by-step noise suppression through a preset sliding window to obtain a noise suppressed image corresponding to the selected multi-color visual image, and then the horizontal feature difference and the vertical feature difference of the noise suppressed image are determined, and the noise suppressed image is converted into a multi-color visual sharpened image according to the horizontal feature difference and the vertical feature difference.

[0046] In this embodiment, the noise suppression is performed step by step on the selected multi-color visual image through a preset sliding window to obtain the noise suppression image corresponding to the selected multi-color visual image in the following manner, namely:

[0047] Suppressing the correlated noise of the selected multi-color visual image according to the sliding window, thereby obtaining a multi-color visual image after suppressing the correlated noise;

[0048] The uncorrelated noise is suppressed on the polychromatic visual image after the correlated noise is suppressed, so as to obtain a noise suppressed image corresponding to the selected polychromatic visual image.

[0049] In specific implementation, first, the sliding window size can be preset through historical experiments and data analysis, so as to suppress the correlated noise of the selected multi-color visual image according to the sliding window. The correlated noise suppression process can be expressed as follows:

[0050]

[0051] in, Represents the pixel point after correlated noise suppression , Indicates that the sliding window contains pixels in the selected multi-color visual image, and They represent the variance and mean of all pixels in the sliding window in the selected multi-color visual image, Represents the minimum value of the constant. By suppressing the correlated noise in the above manner, a multi-color visual image after suppressing the correlated noise can be obtained. It should be noted that due to the correlation between adjacent signals, the readout circuit of the image sensor will interfere with the adjacent signals, thereby introducing stripe noise and color spot noise, namely correlated noise, so it is necessary to suppress the correlated noise; then, the uncorrelated noise usually manifests itself as random noise and uniform noise points, and its sources are mainly dark current and quantization noise. Therefore, it is necessary to suppress the uncorrelated noise of the multi-color visual image after suppressing the correlated noise. The uncorrelated noise suppression process can be expressed as follows:

[0052]

[0053] in, Represents pixel The uncorrelated noise suppression result is Represents the pixel point after correlated noise suppression , Indicates that the sliding window contains pixels in the selected multi-color visual image, Represents a binary noise matrix, which is used to indicate the position of uncorrelated noise in the sliding window. It is obtained by comparing the median error of each pixel in the sliding window with the sliding window. The noise suppression image corresponding to the selected polychromatic visual image can be obtained in the above manner.

[0054] In this embodiment, the horizontal feature difference and the vertical feature difference of the noise suppressed image are determined by taking the pixel standard deviation of the noise suppressed image in the horizontal direction as the horizontal feature difference of the noise suppressed image, and taking the pixel standard deviation of the noise suppressed image in the vertical direction as the vertical feature difference of the noise suppressed image; it should be noted that, in the present application, the horizontal feature difference represents the degree of difference in the features of the noise suppressed image in the horizontal direction, and the vertical feature difference represents the degree of difference in the features of the noise suppressed image in the vertical direction. The horizontal feature difference and the vertical feature difference can help further judge the details and sharpness of the noise suppressed image, especially before the image sharpening process, as a reference for evaluating image features.

[0055] Preferably, in this embodiment, reference Figure 2 As shown in FIG. 1 , this figure is an exemplary flow chart of determining a multi-color visually sharpened image in an embodiment of the present application. In this embodiment, converting the noise suppressed image into a multi-color visually sharpened image according to the horizontal feature difference amount and the vertical feature difference amount can be specifically implemented by the following steps:

[0056] First, in step S21, the aspect ratio is determined by the horizontal feature difference and the vertical feature difference;

[0057] Then, in step S22, the noise suppression image is convolved according to the aspect difference ratio, the horizontal feature difference amount and the vertical feature difference amount, so as to obtain a multi-color visually sharpened image.

[0058] In the specific implementation, first, the aspect difference ratio is determined by the horizontal feature difference and the vertical feature difference, that is, the ratio of the horizontal feature difference to the vertical feature difference is used as the aspect difference ratio, and the aspect difference ratio can be used to measure the shape difference of the Gaussian function; then, the noise suppression image can be convolved according to the aspect difference ratio, the horizontal feature difference and the vertical feature difference, and the convolution process can be represented by the following method:

[0059]

[0060] in, Represents pixel The convolution result is represents the pixel point in the noise suppression image, x and y represent the horizontal and vertical coordinates of the pixel point respectively. and They represent the horizontal feature difference and vertical feature difference of the noise suppressed image respectively, Represents the aspect ratio. The convolution of the noise suppressed image can be completed in the above manner to obtain a multi-color visually sharpened image.

[0061] It should be noted that through the step-by-step noise suppression of the sliding window, the noise can be removed in a targeted manner in the local area. After noise suppression, the image details and texture changes can be quantified by calculating the pixel standard deviation of the image in the horizontal and vertical directions (that is, the horizontal feature difference and the vertical feature difference). Adjusting the degree of sharpening processing by the horizontal feature difference and the vertical feature difference helps to avoid over-sharpening of the image and prevent excessive artifacts or loss of details.

[0062] In step S3, a detail information fusion factor corresponding to the selected multi-color visual image is determined from the multi-color visually sharpened image, and a descriptor is extracted from the multi-color visually sharpened image to obtain a visual feature descriptor corresponding to the selected multi-color visual image.

[0063] Preferably, in this embodiment, reference Figure 3 As shown in FIG. 1 , this figure is an exemplary flow chart of determining the detail information fusion factor corresponding to the selected multi-color visual image in an embodiment of the present application. In this embodiment, the health monitoring coefficient of the smart watch user in a resting state is determined based on the heart rate variability and the pulse variability, which can be specifically implemented by the following steps:

[0064] First, in step S31, a preset local window is obtained;

[0065] Then, in step S32, all local information entropies in the multi-color visually sharpened image are determined based on the local window;

[0066] Finally, in step S33, the detail information fusion factor corresponding to the selected multi-color visual image is determined according to all the local information entropies.

[0067] In specific implementation, first, the size of the local window can be selected based on the resolution of the image and the size of the detail features to be extracted. For example, if the details in the multi-color visual sharpened image are relatively fine, a smaller local window is selected; if the details of the multi-color visual sharpened image are relatively large or complex, a larger local window is selected, which will not be repeated here; then, all local information entropies in the multi-color visual sharpened image can be determined based on the local window, that is, a local window is used to slide in the multi-color visual sharpened image (such as using a sliding window method), and the local information entropy of the multi-color visual sharpened image in each local window area is calculated one by one, and the local information entropy represents the detail intensity and complexity of the corresponding multi-color visual sharpened image area; finally, the detail information fusion factor corresponding to the selected multi-color visual image can be determined based on all the local information entropies, that is, the result after normalizing the mean of all the local information entropies is used as the detail information fusion factor corresponding to the selected multi-color visual image, and the detail information fusion factor represents the fusion weight when fusing the detail information and features of the selected multi-color visual image.

[0068] In this embodiment, descriptor extraction for the multi-color visually sharpened image is performed on the multi-color visually sharpened image by using a recurrent neural network. In specific implementation, the multi-color visually sharpened image is converted into a sequence number suitable for processing by a recurrent neural network. The multi-color visually sharpened image can be divided into grids, and each grid unit is used as a time step of the recurrent neural network to construct a recurrent neural network model. By training the recurrent neural network model, patterns and features in the multi-color visually sharpened image are learned. The recurrent neural network model is trained using labeled data. The trained recurrent neural network model can be used to extract visual descriptors of the multi-color visually sharpened image. The visual descriptor can be used as a visual feature descriptor corresponding to the selected multi-color visual image. The visual feature descriptor represents the image visual features of the selected multi-color visual image.

[0069] It should be noted that the detail information fusion factor is calculated based on the local information entropy of the image, which can effectively reflect the detail complexity of different areas in the image and can be used to measure the importance of the detail information and features of the image. The detail information fusion factor can be used as a weighting coefficient to determine the contribution ratio of different areas in the fusion, and by extracting the visual feature descriptor of the image through a recurrent neural network, high-dimensional and discriminative features can be effectively extracted from the image.

[0070] In step S4, other multi-color visual images are continuously selected to obtain the detail information fusion factor and the corresponding visual feature descriptor corresponding to each multi-color visual image, and the visual feature descriptor corresponding to each multi-color visual image is feature fused based on the corresponding detail information fusion factor to obtain the feature-fused visual feature descriptor, and a multi-color visual enhanced image of the paper to be detected is generated based on the feature-fused visual feature descriptor.

[0071] In specific implementation, other multi-color visual images can be selected, and the multi-color visual images can be processed using the methods in steps S2 to S3 to obtain the detail information fusion factor and the corresponding visual feature descriptor corresponding to each multi-color visual image, which will not be repeated here.

[0072] In this embodiment, the visual feature descriptors corresponding to each multi-color visual image are feature fused based on the corresponding detail information fusion factor, and the visual feature descriptors after feature fusion are obtained in the following manner, namely:

[0073] The detail information fusion factor corresponding to each multi-color visual image and the corresponding visual feature descriptor are multiplied respectively, and the sum of all multiplication results is used as the visual feature descriptor after feature fusion.

[0074] It should be noted that in the present application, the visual feature descriptor after feature fusion is a descriptor that integrates the detail information and features in all polychromatic visual images; through the above method, the visual performance of the detail-rich part of the polychromatic visual image can be enhanced, while the influence of the smooth area can be reduced. This process can effectively enhance the important features in the polychromatic visual image and reduce noise and irrelevant details, thereby improving the final image quality.

[0075] In this embodiment, the multi-color visual enhanced image of the paper to be detected is generated based on the visual feature descriptor after the feature fusion, and the visual feature descriptor after the feature fusion is mapped to the image space by using the reverse mapping technology, so as to obtain the multi-color visual enhanced image of the paper to be detected; in specific implementation, first, the visual feature descriptor after the feature fusion can be input into the corresponding decoder part (such as the deconvolution network) for image space mapping by using the reverse mapping method; then, after the reverse mapping, the image generated by the decoder will be enhanced in terms of details, texture, color, etc., and the network will generate a new image based on the input features, namely the multi-color visual enhanced image of the paper to be detected, which is enhanced in details and also includes optimization processing such as denoising and sharpening.

[0076] It should be noted that, by generating a multi-color visual enhanced image through a feature fusion method based on the detail information fusion factor, the detail information in different multi-color visual images can be selectively fused into the final image. Due to the adaptive adjustment of the detail information fusion factor, the structure of the multi-color visual image (such as shape, contour, etc.) can be maintained, avoiding image distortion due to excessive enhancement, especially when processing complex or high-dimensional images, it can avoid detail loss.

[0077] It can be seen that in the present application, denoising can be effectively performed during the fusion process, and a suitable fusion ratio can be adaptively selected; wherein, through the step-by-step noise suppression of the sliding window, noise can be removed in a targeted manner in the local area, and after noise suppression, by calculating the pixel standard deviation of the image in the horizontal and vertical directions (i.e., the horizontal feature difference and the vertical feature difference), the image details and texture changes can be quantified, and the degree of sharpening processing can be adjusted by the horizontal feature difference and the vertical feature difference, which helps to avoid excessive sharpening of the image; then, the detail information fusion factor can be used as a weighting coefficient to determine the contribution ratio of different regions in the fusion, and by extracting the visual feature descriptor of the image through a recurrent neural network, high-dimensional, discriminative features can be effectively extracted from the image; finally, a multi-color visual enhanced image is generated by a feature fusion method based on the detail information fusion factor, and the detail information in different multi-color visual images can be selectively fused into the final image to avoid image distortion due to excessive enhancement, thereby improving the credibility of paper quality detection.

[0078] In summary, the technical solution adopted in this application can improve the enhancement effect of multi-color visual images during the fusion process to enhance the credibility of paper quality detection.

[0079] Embodiment 2: This application provides a multi-color visual image enhancement system based on image fusion, referring to Figure 4 As shown, this figure is a schematic diagram of an image enhancement system according to this embodiment of the present application, and the image enhancement system includes:

[0080] The image acquisition module 100 is used to acquire all multi-color visual images of the paper to be detected, and select one multi-color visual image from all the multi-color visual images;

[0081] A noise suppression module 200 is used to perform step-by-step noise suppression on the selected multi-color visual image through a preset sliding window to obtain a noise suppressed image corresponding to the selected multi-color visual image, and then determine a horizontal feature difference amount and a vertical feature difference amount of the noise suppressed image, and convert the noise suppressed image into a multi-color visual sharpened image according to the horizontal feature difference amount and the vertical feature difference amount;

[0082] A feature extraction module 300 is used to determine a detail information fusion factor corresponding to a selected multi-color visual image from the multi-color visually sharpened image, extract a descriptor from the multi-color visually sharpened image, and then obtain a visual feature descriptor corresponding to the selected multi-color visual image;

[0083] The image enhancement module 400 is used to continue selecting other multi-color visual images, and then obtain the detail information fusion factor and the corresponding visual feature descriptor corresponding to each multi-color visual image, perform feature fusion on the visual feature descriptor corresponding to each multi-color visual image based on the corresponding detail information fusion factor, obtain the feature-fused visual feature descriptor, and generate a multi-color visual enhanced image of the paper to be detected based on the feature-fused visual feature descriptor.

[0084] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the 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 the 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.

[0085] A person skilled in the art may understand that all or part of the steps in the various methods of the above embodiments may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, the storage medium including 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 programmable 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.

[0086] 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 multi-color visual image enhancement method based on image fusion, characterized in that: The image enhancement method comprises the following steps: Acquire all multi-color visual images of the paper to be detected, and select one multi-color visual image from all the multi-color visual images; Performing step-by-step noise suppression on the selected multi-color visual image through a preset sliding window to obtain a noise suppressed image corresponding to the selected multi-color visual image, and then determining a horizontal feature difference amount and a vertical feature difference amount of the noise suppressed image, and converting the noise suppressed image into a multi-color visual sharpened image according to the horizontal feature difference amount and the vertical feature difference amount; The horizontal feature difference and the vertical feature difference of the noise suppression image are determined by taking the pixel standard deviation of the noise suppression image in the horizontal direction as the horizontal feature difference of the noise suppression image, and taking the pixel standard deviation of the noise suppression image in the vertical direction as the vertical feature difference of the noise suppression image; The method of converting the noise suppressed image into a multi-color visually sharpened image according to the horizontal feature difference and the vertical feature difference specifically includes: Determine the aspect ratio by using the horizontal feature difference and the vertical feature difference; Convolving the noise suppression image according to the aspect difference ratio, the horizontal feature difference amount and the vertical feature difference amount, thereby obtaining a multi-color visually sharpened image; Determining a detail information fusion factor corresponding to a selected multi-color visual image from the multi-color visually sharpened image, performing descriptor extraction on the multi-color visually sharpened image, and then obtaining a visual feature descriptor corresponding to the selected multi-color visual image; Continue to select other multi-color visual images, and then obtain the detail information fusion factor and the corresponding visual feature descriptor corresponding to each multi-color visual image, based on the corresponding detail information fusion factor, perform feature fusion on the visual feature descriptor corresponding to each multi-color visual image to obtain the feature-fused visual feature descriptor, and generate a multi-color visual enhanced image of the paper to be detected based on the feature-fused visual feature descriptor.

2. The multi-color visual image enhancement method based on image fusion as claimed in claim 1, characterized in that: All multi-color visual images of the paper to be inspected are acquired through the image sensor.

3. The multi-color visual image enhancement method based on image fusion as claimed in claim 1, characterized in that: The noise suppression of the selected multi-color visual image is performed step by step through a preset sliding window to obtain a noise suppression image corresponding to the selected multi-color visual image, specifically including: Suppressing the correlated noise of the selected multi-color visual image according to the sliding window, thereby obtaining a multi-color visual image after suppressing the correlated noise; The uncorrelated noise is suppressed on the polychromatic visual image after the correlated noise is suppressed, so as to obtain a noise suppressed image corresponding to the selected polychromatic visual image.

4. The multi-color visual image enhancement method based on image fusion as claimed in claim 1, characterized in that: Determining the detail information fusion factor corresponding to the selected multi-color visual image from the multi-color visual sharpened image specifically includes: Get the preset local window; determining all local information entropies in the polychromatic visually sharpened image based on the local window; The detail information fusion factor corresponding to the selected multi-color visual image is determined according to all local information entropies.

5. The multi-color visual image enhancement method based on image fusion as claimed in claim 1, characterized in that: The descriptor extraction for the multi-color visually sharpened image is performed by extracting the descriptor for the multi-color visually sharpened image through a recurrent neural network.

6. The multi-color visual image enhancement method based on image fusion as claimed in claim 1, characterized in that: Based on the corresponding detail information fusion factor, the visual feature descriptors corresponding to each multi-color visual image are fused to obtain the visual feature descriptors after feature fusion, which specifically include: The detail information fusion factor corresponding to each multi-color visual image and the corresponding visual feature descriptor are multiplied respectively, and the sum of all multiplication results is used as the visual feature descriptor after feature fusion.

7. The multi-color visual image enhancement method based on image fusion as claimed in claim 1, characterized in that: Generating a multi-color visually enhanced image of the paper to be detected based on the visual feature descriptor after the feature fusion is to map the visual feature descriptor after the feature fusion to the image space using a reverse mapping technology, thereby obtaining a multi-color visually enhanced image of the paper to be detected.

8. A multi-color visual image enhancement system based on image fusion, used to execute a multi-color visual image enhancement method based on image fusion as claimed in any one of claims 1 to 7, characterized in that: The image enhancement system comprises: An image acquisition module is used to acquire all multi-color visual images of the paper to be detected and select one multi-color visual image from all the multi-color visual images; A noise suppression module is used to perform step-by-step noise suppression on the selected multi-color visual image through a preset sliding window to obtain a noise suppressed image corresponding to the selected multi-color visual image, and then determine a horizontal feature difference amount and a vertical feature difference amount of the noise suppressed image, and convert the noise suppressed image into a multi-color visual sharpened image according to the horizontal feature difference amount and the vertical feature difference amount; A feature extraction module, used to determine a detail information fusion factor corresponding to a selected multi-color visual image from the multi-color visually sharpened image, extract a descriptor from the multi-color visually sharpened image, and then obtain a visual feature descriptor corresponding to the selected multi-color visual image; The image enhancement module is used to continue selecting other multi-color visual images, and then obtain the detail information fusion factor and the corresponding visual feature descriptor corresponding to each multi-color visual image, perform feature fusion on the visual feature descriptor corresponding to each multi-color visual image based on the corresponding detail information fusion factor, obtain the feature-fused visual feature descriptor, and generate a multi-color visual enhanced image of the paper to be detected based on the feature-fused visual feature descriptor.

Images