Terahertz near-field image splicing method based on combination of edge features and information entropy

Abstract

The invention discloses a terahertz near-field image splicing method based on the combination of edge features and information entropy, and the method comprises the steps: firstly, selectively calculating the edge features of an image to improve the processing speed, secondly, obtaining the corresponding position relation of a to-be-spliced image based on the edge features by applying a matching algorithm, obtaining the overlapping region of the to-be-spliced image, and finally, carrying out the splicing of the to-be-spliced image. Information entropy evaluation indexes are introduced, high-quality overlapped areas are screened and reserved to enhance the imaging quality, the steps are repeated, pairwise matching of multiple adjacent images is completed in sequence, and splicing of multiple small-range images is expanded into splicing of multiple large-range images. According to the method, the imaging range is rapidly expanded, the imaging definition is improved, and the method has the advantages of being easy and convenient to implement and high in efficiency.

Description

Technical Field

[0001] The present invention belongs to the field of terahertz near-field system imaging, and in particular relates to a terahertz near-field image stitching method based on the combination of edge features and information entropy. Background Art

[0002] The terahertz near-field imaging system is an advanced imaging technology that combines the characteristics of terahertz waves to achieve non-destructive high-resolution imaging of samples. The system is mainly composed of a terahertz source, a terahertz optical path, a parabolic mirror, a probe platform, etc. The terahertz source is the core part of the system, which can emit terahertz waves with strong penetration, which are transmitted through a specific terahertz optical path and focused on the probe platform through a parabolic mirror. The sample is placed under the probe, and the terahertz wave interacts with it. The reflected wave carries the near-field information of the sample. The signal processing system collects the terahertz waves reflected by the probe, and completes the internal imaging of the sample through processing and analysis. The terahertz near-field imaging system can achieve nanometer-level imaging accuracy, which is very effective for observing the details of microstructures and biological tissues. This imaging technology has broad application prospects in biomedicine, materials science, semiconductor industry and other fields, especially in situations requiring high resolution and non-destructive testing.

[0003] However, the number of scanning points in the near-field imaging system is fixed, which means that when the system scanning range is reduced, the spacing between the scanning points is correspondingly reduced. Although the corresponding scanning data accuracy becomes higher at this time, the reduction in the scanning range reduces the robustness of the calculation results. Therefore, how to obtain image data with a larger scanning range and good quality has become a problem that needs to be solved urgently. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a terahertz near-field image stitching method based on edge features and information entropy. After selecting a key scanning area, the near-field imaging device moves the probe in the selected area to collect data. In order to obtain image data of a larger scanning range while maintaining high precision, the method first performs high-precision small-range scanning to obtain a small-range image, then stitches the small-range image, extracts the edge features of the small-range image, and determines the relative position and direction of each small-range image through feature matching technology. Based on this information, the small-range image is stitched into a large-range image, and finally the above process is repeated to achieve stitching of multiple small-range images. The stitching of multiple images obtains larger scanning range data, solving the problem of small scanning range. During the acquisition process, there will be overlapping areas between two adjacent images. When stitching, it is necessary to select the overlapping areas. For the overlapping areas of the two input images, information entropy is introduced to evaluate the image quality of the overlapping areas, and the overlapping areas with high information entropy evaluation are retained, thereby ensuring the high quality of the overlapping areas of the stitched images.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0006] The terahertz near-field image stitching method based on the combination of edge features and information entropy has the following steps:

[0007] Step 1: collect data of imaging samples based on the terahertz near-field imaging system to obtain a small-range high-resolution imaging image of multiple imaging samples;

[0008] Step 2: Calculate the edge area features of two adjacent small-scale high-resolution imaging images based on the SIFT feature method to obtain a feature description and a feature point position set of each small-scale high-resolution imaging image;

[0009] Step 3: Use the K nearest neighbor matching algorithm to match the feature descriptions and feature point positions of the two adjacent small-range high-resolution imaging images, obtain the position correspondence between the two small-range high-resolution imaging images, and perform splicing;

[0010] Step 4: Calculate the information entropy for the overlapping area of ​​each stitching result image, and select the stitching result image with a high information entropy value as the final stitching result image of the two small-range high-resolution imaging images;

[0011] Step 5: stitch the final stitching result image and the next small-scale high-resolution imaging image to be stitched together according to steps 2 to 4 to obtain a new final stitching result image, until the last small-scale high-resolution imaging image to be stitched together is stitched into a large-scale image.

[0012] Furthermore, the edge area is 20% of the area around the small-scale high-resolution imaging image.

[0013] Furthermore, the information entropy in step 4 The calculation formula is:

[0014] ;

[0015] in, Represents the gray value of the pixel, It represents the proportion of pixels with gray value i in the image, obtained by gray histogram statistics.

[0016] On the other hand, the present invention also provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned terahertz near-field image stitching method based on the combination of edge features and information entropy.

[0017] In a third aspect, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enables the processor to implement the aforementioned terahertz near-field image stitching method based on the combination of edge features and information entropy.

[0018] The beneficial effects of the present invention are:

[0019] The present invention optimizes the data processing strategy and only performs feature calculation on the edge area of ​​the image, thereby improving the calculation efficiency and laying the foundation for subsequent image processing. The introduction of the KNN algorithm helps to accurately identify and correspond to key feature points in the image. Information entropy can quantify the amount of information of each pixel in the image. Using information entropy as an indicator for evaluating images can improve the quality of spliced ​​images and reduce the impact of noise. Image fusion is performed on the selected areas with high information entropy to generate the final high-quality imaging results, thereby ensuring the accuracy and reliability of the imaging results and expanding the application scope of the terahertz near-field imaging system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of the terahertz near-field image stitching method based on the combination of edge features and information entropy of the present invention;

[0021] Figure 2 Schematic diagram of collecting sample data according to an embodiment of the present invention, wherein (a) is a stained sample image collected under a microscope, and (b) is a detection image of an unstained sample image by a near-field imaging system;

[0022] Figure 3 Schematic diagram of the principle of calculating image edge region features, where (a) and (b) are two images to be stitched;

[0023] Figure 4 It is a schematic diagram of the principle of edge region feature matching graph;

[0024] Figure 5 The overlapping areas of adjacent images, where (a)(b), (c)(d), (e)(f) are the imaging results of two acquisitions in three different areas respectively;

[0025] Figure 6 The comparison diagram of the stitching results before and after the introduction of the information entropy evaluation index, where (a) is the stitching result diagram of the weighted fusion of the overlapping areas, and (b) is the stitching result diagram after the introduction of the information entropy stitching index. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0027] like Figure 1 As shown, the present invention provides It's The terahertz near-field image stitching method based on the combination of edge features and information entropy specifically includes the following steps:

[0028] Example

[0029] The present invention is described below using a thyroid papillary carcinoma tissue image as a sample image. The entire image stitching process is as follows: Figure 1 shown.

[0030] Step 1: Figure 2 (a) shows a sample of thyroid papillary carcinoma tissue image stained under a microscope. The terahertz near-field imaging system is used to collect data on the tissue sample and obtain multiple small-scale high-resolution images of the thyroid tissue. Figure 2 as shown in (b).

[0031] Step 2: Calculate the edge area features of the small-scale high-resolution imaging image. The specific operation is as follows: Figure 3 (a) and (b) show two small-scale high-resolution images collected by the adjacent time terahertz near-field imaging system. The size is 256x256, and 20% of the area on the edge of the input image is selected, that is, Figure 3 As shown in the outer light-colored areas of (a) and (b), assuming that the upper left corner of the input image is the coordinate origin, the horizontal coordinate x and the vertical coordinate y are the row and column of the image respectively, and the four edge side feature areas can be represented as a rectangular box (x_min, y_min, x_max, y_max), where (x_min, y_min) is the coordinate point of the upper left corner of the rectangular box, and (x_max, y_max) is the coordinate point of the lower right corner of the rectangular box. The positions of the four edge side areas are: (0, 0, 256, 51), (0, 201, 256, 256), (0, 51, 51, 201), (201, 51, 256, 201). The present invention selects SIFT features (Scale-Invariant Feature Transform) to extract edge side features of image (a) and image (b) to obtain feature descriptions and feature point position sets for each small-scale high-resolution imaging image.

[0032] Step 3: Determine feature matching coordinates: Use the K nearest neighbor (KNN) matching algorithm to match the feature descriptions of the two adjacent small-scale high-resolution imaging images obtained in step 2 to determine the positional correspondence between the two images and achieve the splicing of the two images. The positional correspondence of the images can be shown as follows: Figure 3 In the actual experiment, due to too many matching feature points, the plotting on the image appears crowded. In order to improve the visualization effect, only the first 50 best matching points are selected for plotting, as shown in Figure 4 As shown in the figure, the correspondence between four groups of adjacent captured images is shown. The position of the feature points in the edge area is indicated by a circle in the figure, and the corresponding relationship between the feature descriptions of two adjacent images is shown by the connecting lines. In order to verify the impact of the extracted edge features on the speed, four common spliced ​​data sets were selected for testing. Table 1 summarizes the performance of the four data sets. The data shows that the feature extraction and matching speed has been optimized.

[0033] Table 1

[0034] Dataset Extract global feature(s) Extract edge feature(s) zirconREM 3.429 2.969 zirconBSE 0.851 0.561 zirconTEM 4.350 2.840 zirconCL 5.799 3.643

[0035] Step 4: Use information entropy index to evaluate the imaging quality of the overlapped area: Since the actual system has noise, the quality of the collected image is unstable, such as Figure 5 As shown in the figure, (a)(b), (c)(d), (e)(f) are the imaging results of two acquisitions in three different areas. For example, although (a)(b) is the same area, the stitching quality of the images acquired twice is different. Therefore, after determining the positional relationship of the two images in step 3, it is necessary to select the overlapping areas of multiple sets of stitched images, and select the ones with high imaging quality as the final stitching result. Information entropy is introduced as an evaluation indicator. Information entropy The calculation formula is:

[0036] ;

[0037] in, Represents the gray value of the pixel, It represents the proportion of pixels with gray value i in the image, obtained by gray histogram statistics.

[0038] Image-based information entropy is a statistical form of features that reflects the average amount of information in an image. It indicates the amount of information contained in the aggregated features of the grayscale distribution in an image. It can be used to measure the imaging quality of overlapping areas. Figure 5 The information entropy values ​​of the three groups of pictures (a)(b), (c)(d), and (e)(f) are calculated, and their values ​​are as follows: (a) is 19.6, (b) is 9.2, (c) is 13.7, (d) is 10.3, (e) is 12.6, and (f) is 29.2. It can be seen that the clear image corresponds to a high information entropy value.

[0039] Therefore, the information entropy of the overlapping area is calculated, and the stitching imaging result with a high information entropy value is selected as the final stitching result image of the two images.

[0040] Step 5. After completing the stitching of two adjacent images in step 4 and obtaining the final stitching result image, stitch the final stitching result image and the next image to be stitched according to steps 2 to 4 to obtain a new final stitching result image. Repeat this process until the last image to be stitched, and complete the pairwise matching of multiple adjacent images. Multiple small-range images are stitched and expanded into a large-range image.

[0041] Figure 6 To compare the stitching effects before and after the introduction of the information entropy index, (a) is the stitching using a 50% weighted fusion method for the overlapping area, and (b) is the final stitching result after the introduction of the information entropy index. Compared with the box marked area in (a), the image in (b) is clearer and preserves a lot of subtle information. This step ensures the accuracy and reliability of the imaging results, while expanding the application range of terahertz near-field imaging systems.

[0042] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A terahertz near-field image stitching method based on edge features and information entropy, characterized in that: Here are the steps: Step 1: collect data of imaging samples based on the terahertz near-field imaging system to obtain a small-range high-resolution imaging image of multiple imaging samples; Step 2: Calculate the edge area features of two adjacent small-scale high-resolution imaging images based on the SIFT feature method to obtain a feature description and a feature point position set of each small-scale high-resolution imaging image; Step 3: Use the K nearest neighbor matching algorithm to match the feature descriptions and feature point positions of the two adjacent small-range high-resolution imaging images, obtain the position correspondence between the two small-range high-resolution imaging images, and perform splicing; Step 4: Calculate the information entropy for the overlapping area of ​​each stitching result image, and select the stitching result image with a high information entropy value as the final stitching result image of the two small-range high-resolution imaging images; Step 5: stitch the final stitching result image and the next small-scale high-resolution imaging image to be stitched together according to steps 2 to 4 to obtain a new final stitching result image, until the last small-scale high-resolution imaging image to be stitched together is stitched into a large-scale image.

2. The terahertz near-field image stitching method based on edge feature and information entropy according to claim 1 is characterized in that: The edge area is 20% of the area around the small-scale high-resolution imaging image.

3. The terahertz near-field image stitching method based on edge feature and information entropy according to claim 1 is characterized in that: The information entropy in step 4 The calculation formula is: ， in, Represents the gray value of the pixel, It represents the proportion of pixels with gray value i in the image, which is obtained by gray histogram statistics.

4. An electronic device, characterized in that: include: one or more processors; A memory for storing one or more programs; Among them, when one or more programs are executed by the one or more processors, the one or more processors implement the terahertz near-field image stitching method based on the combination of edge features and information entropy as described in any one of claims 1-3.

5. A computer-readable storage medium, characterized in that: Executable instructions are stored thereon, and when the instructions are executed by the processor, the processor can implement the terahertz near-field image stitching method based on the combination of edge features and information entropy as described in any one of claims 1-3.

Images