Feature point calculation, description and matching method applied to multi-channel image data

Abstract

The application provides a feature point calculation, description and matching method applied to multi-channel image data, which extracts stable features of an image by establishing a multi-channel color change scale space and judging the extreme value of pixels in the scale space. Furthermore, the features are used to calculate the matching of the same features between different images, so that the feature point calculation, description and matching of multi-channel images are realized, and a more accurate and comprehensive method is provided for image processing and analysis. The application can use pixel information in a small range of neighborhood to complete the description and calculation of feature points, and effectively improves the accuracy of the feature point matching structure. The implementation of the method will bring new breakthrough and progress to the field of image processing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a feature point calculation, description and matching method applied to multi-channel image data. BACKGROUND

[0002] Most of the traditional image processing techniques take gray-scale images as the main processing object, which is mainly because the data volume of the gray-scale images is smaller and the processing speed is relatively faster. However, with the continuous maturity of color image sensor technology, high-resolution color images can provide more abundant structural and scene detail information. However, when processing color images, the traditional image processing method usually still converts them into gray-scale images, and this conversion process inevitably loses a large amount of structural and scene detail information.

[0003] When calculating the structural and scene feature points in the image, more extensive pixel information needs to be used to ensure the accuracy of the calculation and the precision of the description. This further puts higher requirements on the static state of the structure and the scene. If the state of the structure or the scene changes during the image sampling process, the description information of the feature points will also change, which will reduce the robustness and accuracy of the feature point extraction and matching. SUMMARY

[0004] In order to overcome the shortcomings of the prior art, the purpose of the present application is to provide a feature point calculation, description and matching method applied to multi-channel image data.

[0005] To achieve the above purpose, the present application provides the following scheme:

[0006] A feature point calculation, description and matching method applied to multi-channel image data, comprising:

[0007] obtaining multi-channel image data, a pre-designed calculation neighborhood, a feature point calculation neighborhood and a matching calculation neighborhood; each pixel point in the multi-channel image data comprises a plurality of color channels; the size of the multi-channel image data is [m x n]; m and n respectively represent the number of pixels in each direction of the image on the pixel plane; the pre-designed calculation neighborhood is α pixel units; the feature point calculation neighborhood is β pixel units; and the matching calculation neighborhood is γ pixel units;

[0008] taking a square region with a side length of 2α as the center of each coordinate point from (α, α) coordinate point to [(m-α), (n-α)] coordinate point, calculating the difference value of each color channel of all pixel points in the square region of each coordinate point and each color channel of the pixel of each coordinate point, and obtaining all coordinate point pixel 2α x 2α-1 value;

[0009] The square region with a side length of 2a is taken with the (a+β, a+β) coordinate point as the center, and the difference value of each color channel of all pixel points in the square region of each coordinate point and each color channel of the pixel of each coordinate point is calculated to obtain 2a*2a-1 pixel points of all coordinate points values

[0010] The (a+β, a+β) point is taken as the center, and β is taken as the radius, and whether all values of the (a+β, a+β) point are extreme values in the neighborhood is verified, if yes, the point is defined as a feature point, if not, the point is not a feature point

[0011] All feature points of two images image_1 and image_2 are calculated respectively to obtain the feature distribution of the two image data under the calculation mode

[0012] The feature point pixel coordinates of any point in image_1 are taken as the center, and γ is taken as the radius, and the difference values of all values of the feature points and all feature points in image_2 within the range of γ are calculated respectively, if the difference values are lower than the threshold ζ, it indicates that the two are the same named points, and the matching calculation of the feature point is completed, if there is no feature point in image_2 within the threshold ζ, that is, there is no same named point in image_2.

[0013] According to the specific embodiments provided by the application, the following technical effects are disclosed:

[0014] The application provides a feature point calculation, description and matching method applied to multi-channel image data, which can utilize pixel information in a small range neighborhood to complete the description and calculation of feature points, and effectively improves the accuracy of the feature point matching structure. The implementation of the method will bring new breakthroughs and progress to the field of image processing. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0016] Figure 1 The method flowchart provided by the embodiments of the present application. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the scope of the present application.

[0018] The present application aims to provide a feature point calculation, description and matching method applied to multi-channel image data, which can effectively improve the accuracy of feature point matching structure.

[0019] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0020] Figure 1 The method flowchart provided by the embodiments of the present application is shown in Figure 1 The present application provides a feature point calculation, description and matching method applied to multi-channel image data, which includes:

[0021] Step 100: obtaining multi-channel image data, pre-designing calculation neighborhood, feature point calculation neighborhood and matching calculation neighborhood; each pixel point in the multi-channel image data includes multiple color channels; the size of the multi-channel image data is [m x n]; m and n respectively represent the number of pixels in each direction of the image on the pixel plane; the pre-designing calculation neighborhood is alpha pixel units; the feature point calculation neighborhood is beta pixel units; and the matching calculation neighborhood is gamma pixel units;

[0022] Step 200: taking the square region with a side length of 2a as the center in turn from the (a, a) coordinate point to the [(m-a), (n-a)] coordinate point, calculating the difference value of each color channel of all pixel points in the square region of each coordinate point and each color channel of the pixel of each coordinate point, and obtaining 2a x 2a-1 values of all coordinate points pixels;

[0023] Step 300: taking the square region with a side length of 2a as the center in turn from the (a+β, a+β) coordinate point to the [(m-a-β), (n-a-β)] coordinate point, calculating the difference value of each color channel of all pixel points in the square region of each coordinate point and each color channel of the pixel of each coordinate point, and obtaining 2a x 2a-1 values of all coordinate points pixels;

[0024] Step 400: taking the (a+β, a+β) point as the center and β as the radius, verifying all ​​whether the value is a local extremum, if yes, the point is defined as a feature point, if no, the point is not a feature point;

[0025] Step 500: calculate all feature points of two images image_1 and image_2 respectively; obtain the feature distribution of two image data under the calculation mode;

[0026] Step 600: take the feature point pixel coordinates of any point in image_1 as the center and γ as the radius, respectively calculate the difference value of the feature point all values and all feature points in image_2 within the range of γ, if the difference value is lower than the threshold ζ, it means that the two are the same name points, and the matching calculation of the feature point is completed, if there is no feature point in image_2 within the threshold ζ, that is, there is no same name point in image_2.

[0027] Further, the technical route of the embodiment is as follows:

[0028] Step 1: based on multi-channel image [m x n] data (m, n respectively represent the number of pixels in each direction on the pixel plane), each pixel point in the image contains color channel {channel1, channel2, …, channel n}, count n channels;

[0029] Step 2: set the calculation neighborhood as α pixel units, that is, start calculating from the pixel point at (α, α) coordinates, row by row, column by column, and point by point, until the pixel point at [(m-α), (n-α)] coordinates;

[0030] Specifically, because the image data is represented in a row-column form in the computer, because the description of a pixel point needs to be calculated with the surrounding α units of pixels, therefore, the picture is taken as the upper left corner as the starting origin, the pixel point data from the top down α rows, and from the left to right α columns is not enough neighborhood pixels to represent, therefore, the pixel point from the top down α units and from the left to right α units is the first calculation pixel, the specific calculation is step three, here only represents a macroscopic calculation process from the starting pixel point to the ending pixel point, the coordinate of the last pixel point is α units from the bottom up and α units from the right to the left, at this coordinate point.

[0031] Step 3: take the (α, α) coordinate point as the center and a square region with a side length of 2α, calculate the difference value of each color channel of all pixel points in the calculation region and the color channel of the (α, α) coordinate point pixel, obtain 2α x 2α-1 values at the (α, α) coordinate point;

[0032] Step 4: repeating step 3 from (a, a) to get all coordinate points of 2a x 2a-1 pixels in the rectangular region from (a, a) to [(m-a), (n-a)] value;

[0033] Step 5: further defining the feature point calculation neighborhood as β pixel units, i.e., starting from the pixel point at the (a+β, a+β) coordinate, performing calculation point by point row by row and column by column until the pixel point at the [(m-a-β), (n-a-β)] coordinate.

[0034] Specifically, in addition to all the pixel points in the middle part having been described and calculated in steps 2 and 3 through the neighborhood of a pixels from top to bottom a rows, from left to right a columns, from bottom to top a rows, and from right to left a columns, whether a feature point is a reliable feature point needs to be judged, which is whether it is a prominent point in the neighborhood range, which is determined by the extreme value calculation in step 6. In step 2, only the description information of the middle part of the pixel points is obtained, but it needs to be compared with the neighborhood β range. Therefore, in the middle pixel region, continue to count β rows from top to bottom, β columns from right to left, β rows from bottom to top, and β columns from left to right, and take the remaining middle part of the pixels to perform the extreme value calculation in step 6. Through steps 6 and 7, the specific feature points are obtained.

[0035] Step 6: taking the (a+β, a+β) point as the center and β as the radius to check whether all values of the (a+β, a+β) point are extreme values in the neighborhood, if yes, the point is defined as a feature point, if not, the point is not a feature point.

[0036] Step 7: respectively calculating all the feature points of two images image_1 and image_2 to obtain the feature distribution of the two image data under the calculation mode.

[0037] Step 8: further defining the matching calculation field γ, taking the pixel coordinate of any feature point p0 in image_1 as the center and γ as the radius to respectively calculate the difference values of all values of p0 and all feature points in the γ range in image_2, if the difference values are lower than the threshold ζ, it indicates that the two are the same named points, and the matching calculation of the feature point is completed, if there is no feature point in the γ range in image_2 within the threshold ζ, i.e., there is no same named point of p0 in image_2.

[0038] Step 9: repeating step 8 until the matching calculation of all the feature points of image_1 in image_2 is completed.

[0039] The beneficial effects of the present application are as follows:

[0040] The application provides a feature point calculation, description and matching method based on multi-channel images.

[0041] The various embodiments are described in a progressive manner in the specification, and each embodiment focuses on the difference from other embodiments, and the same or similar parts between the various embodiments can be referred to each other.

[0042] The principles and implementation manners of the application are described by using specific examples in the specification, and the above description of the examples is only used to help understand the method and core idea of the application; meanwhile, according to the idea of the application, the specific implementation manners and application ranges can be changed by the person skilled in the art. In conclusion, the content of the specification should not be understood as the limitation of the application.

Claims

1. A method for feature point calculation, description, and matching applied to multi-channel image data, characterized in that, Feature point description and calculation are performed using pixel information within a small neighborhood, including: The system acquires multi-channel image data, a preset computational neighborhood, a feature point computational neighborhood, and a matching computational neighborhood. Each pixel in the multi-channel image data includes multiple color channels. The size of the multi-channel image data is [m×n], where m and n represent the number of pixels in each direction on the pixel plane. The preset computational neighborhood is α pixel units; the feature point computational neighborhood is β pixel units; and the matching computational neighborhood is γ pixel units. Taking the coordinates from (α, α) to [(m-α), (n-α)] as centers, take a square region with a side length of 2α. Calculate the difference values ​​between each color channel of all pixels within the square region at each coordinate point and the color channels of all pixels at each coordinate point, resulting in 2α × 2α - 1 pixels for all coordinate points. value; Taking the coordinates from (α+β, α+β) to [(m-α-β), (n-α-β)] as centers, a square region with a side length of 2α is drawn. The difference values ​​of each color channel of all pixels within this square region are calculated between the difference values ​​of the pixels at each coordinate point and the difference values ​​of the pixels at each coordinate point. This yields 2α × 2α - 1 pixels at all coordinate points. value; With the point (α+β, α+β) as the center and β as the radius, test all points (α+β, α+β). If the value is an extreme value in the neighborhood, then this point is defined as a feature point; otherwise, this point is not a feature point. Calculate all feature points in two images, image_1 and image_2, respectively; obtain the feature distribution of the two image data under this calculation mode; Using the pixel coordinates of any feature point in image_1 as the center and γ as the radius, calculate all the feature points in the image_1. The value is the difference between the value and all feature points within the γ range in image_2. If the difference is less than the threshold ζ, it means that the two are the same point and the matching calculation of the feature point is completed. If there are no feature points within the threshold ζ range in image_2, that is, no feature point has no corresponding point in image_2.

Images