Vein image enhancement method based on multi-channel filter

Abstract

The invention relates to the technical field of image enhancement, in particular to a vein image enhancement method based on a multi-channel filter, which comprises the following steps of: performing vein information detection by adopting a multi-channel uniform and symmetrical filter with four orientations, determining parameters of the filter according to vein changes of diameters and diffusion modes, and then performing two-dimensional image operation. The vein information is enhanced in different directions, and finally the filtered image is reconstructed to obtain an enhanced vein image. Experimental results show that the method can effectively improve the contrast of the vein image, especially for a low-quality vein image.

Description

Technical Field

[0001] The present invention relates to the technical field of image enhancement, and in particular to a vein image enhancement method based on a multi-channel filter. Background Art

[0002] Veins are located below the epidermis, so visible light cannot image them. In medical applications, NIR (near infrared) lamps are often used for vein imaging because radiation with a wavelength of 760-850nm can be absorbed by deoxyhemoglobin in veins. Vein images always have low contrast due to background noise and intensity fluctuations. Many techniques have been used to enhance vein images. However, current methods generally perform poorly in improving image contrast when image quality varies from person to person. Extracting reliable vein information in real applications is a huge challenge.

[0003] Since the diameter and expansion behavior of veins on the fingers vary greatly, it is necessary to explore characterization methods that can conveniently describe the changes in vein orientation and diameter. Summary of the invention

[0004] In order to enhance the vein image, the present invention provides a vein image enhancement method.

[0005] A vein image enhancement method based on a multi-channel filter comprises the following steps:

[0006] Step 1, obtaining a vein image;

[0007] Step 2, the filter is,

[0008]

[0009] where k (=1, 2, 3, 4) is the channel index, θk (=kπ / 4) and f k denote the orientation and center frequency of the uniform symmetrical filter respectively; ρ = 1 / (2πσ x σ y ), ^j = √-1, θ is the direction of the filter, f0 represents the center frequency of the filter, σ x and σ y are the two axes of the elliptical Gaussian envelope, x θ and θ is the rotated version of the filter's coordinates (x, y);

[0010] Step 3: Obtain the filtered vein image rk(x, y) by convolving the input vein image I(x, y) with the uniform symmetric filter Gek(x, y) at the Kth channel,

[0011]

[0012] Where * indicates that the convolution is in two dimensions, and the directions of the uniform and symmetric filters are 0°, 45°, 90°, and 135° respectively;

[0013] Step 4, image reconstruction, the four filtered images of the filter are at each point pi(x i ,y i ) to form a 4-dimensional array S i , assuming that R(x, y) represents the reconstructed image, the reconstruction rule is defined as:

[0014]

[0015] Where N is the total number of pixels of the vein image.

[0016] The beneficial effects of the present invention are: using a multi-channel uniform symmetric filter with four orientations to detect vein information, determining the parameters of the filter according to the changes in the diameter and diffusion mode of the vein, then performing a two-dimensional image operation to enhance the vein information in different directions, and finally reconstructing the filtered image to obtain an enhanced vein image. Experimental results show that this method can effectively improve the contrast of the vein image, especially for low-quality vein images. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is the venous network reconstruction diagram of the present invention.

[0018] Figure 2 is a filtered image with 4 additional orientations.

[0019] Figure 3 The reconstruction result (a) corresponds to Figure 1 The reconstruction result (b) corresponds to Figure 2 The reconstruction result (c) corresponds to Figure 2 and Figure 1 Results for 8 orientations.

[0020] Figure 4 3D graph representation Figure 3 (a) and Figure 3 The error between the reconstructed images in (c).

[0021] Figure 5 Comparison of vein image enhancement DETAILED DESCRIPTION

[0022] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] The present invention provides a vein image enhancement method based on a multi-channel filter, comprising the following steps:

[0024] Step 1, obtaining a vein image;

[0025] Step 2, the filter is,

[0026]

[0027] where k (=1, 2, 3, 4) is the channel index, θk (=kπ / 4) and f k denote the orientation and center frequency of the uniform symmetrical filter respectively; ρ = 1 / (2πσ x σ y ), θ is the direction of the filter, f0 is the center frequency of the filter, and σ x and σ y are the two axes of the elliptical Gaussian envelope, x θ and θ is the rotated version of the filter's coordinates (x, y);

[0028] Step 3, obtain the filtered vein image rk(x, y) by convolving the input vein image I(x, y) with the uniform symmetric filter Gek(x, y) at the Kth channel,

[0029]

[0030] Where * indicates that the convolution is in two dimensions, and the directions of the uniform and symmetric filters are 0°, 45°, 90°, and 135° respectively;

[0031] Step 4, image reconstruction, the four filtered images of the filter are at each point pi(x i ,y i ) to form a 4-dimensional array S i , assuming that R(x, y) represents the reconstructed image, the reconstruction rule is defined as:

[0032]

[0033] Where N is the total number of pixels of the vein image.

[0034] The two-dimensional filter is represented as a Gaussian function modulated by a complex sinusoidal signal. It is defined as follows:

[0035]

[0036] in:

[0037]

[0038] ρ=1 / (2πσ x σ y ), θ is the orientation of the filter, f0 represents the center frequency of the filter, σx and σy are the two axes of the elliptical Gaussian envelope, and xθ and yθ are the rotated versions of the coordinates (x, y) of the filter.

[0039] The present invention adopts a uniform filter, which is suitable for ridge detection in an image. Since the texture of the NIR imaging mode always presents a dark ridge structure, and there are great differences in the diameter and propagation mode of the finger, a multi-channel uniform symmetric filter is suitable for the detection of vein information. The uniform filter is expressed as:

[0040]

[0041] where k (=1, 2, 3, 4) is the channel index, θk (=kπ / 4) and f k They represent the orientation and center frequency of the uniform symmetric filter respectively.

[0042] The filtered vein image rk(x,y) is obtained by convolving the input vein image I(x,y) with a uniform (symmetric) filter Gek(x,y) at the Kth channel.

[0043]

[0044] Where * indicates that the convolution is in two dimensions. Four filtered images are generated by a uniform symmetric filter with four channels, so that the vein information is successfully enhanced in four directions. The four filtered images are then reconstructed to obtain a comprehensive vein network.

[0045] Image reconstruction, the output of the filter forms a 4-dimensional array Si at each point pi(xi, yi) in the image I(x, y). Assuming R(x, y) represents the reconstructed image, the reconstruction rule is defined as:

[0046]

[0047] Where N is the total number of pixels of the vein image.

[0048] Figure 1 This is the venous network reconstruction map of the present invention, and the 3D map corresponds to the 3D distribution of rk.

[0049] The proposed reconstruction rule can effectively obtain The best response of a pixel in a certain direction. Figure 1 The left side of is shown, where the valleys in the 3D image correspond to the ridges in the image plane. Figure 1 As shown on the right, the reconstructed vein image R(x, y) can be obtained reliably and effectively using the reconstruction rule.

[0050] Typically, reconstructing veins without structural distortion depends on the sensitivity of the filter. Therefore, for vein textures, the parameters of the filter determine its optimal output.

[0051] When the filter period (Tk=1 / fk) is designed to be twice the width of the vein and the direction is orthogonal to the orientation of the vein, the filter response is optimal, which can effectively mine more vein information and reduce computational costs.

[0052] The present invention can also use more orientations of the filter to mine more vein information. However, selecting more than 8 directions will generate more redundant information, which will greatly reduce the performance of the filter.

[0053] When 8 directions are selected, the image reconstruction with 8 orientations (θk=kπ / 8, k=1, ..., 8) is almost the same as that with 4 orientations (θk=kπ / 4, k=1, ...), Figure 3 (a) and Figure 3 The difference between the reconstructed images in (c) is almost zero in the main vein regions, except for some smaller patches. Figure 2 The filtered image has 4 additional orientations, which are 22.5°, 67.5°, 112.5° and 157.5°. A lot of capillary information is lost. The reconstructed image is as follows: Figure 3 (b) shows that the vein information extracted from the four directions of 22.5°, 67.5°, 112.5° and 157.5° is redundant. Therefore, selecting four directions: 0°, 45°, 90°, and 135° can reduce the computational complexity.

[0054] Figure 5 Comparison of vein image enhancement. Figure 5 (a) is the original vein image, Figure 5 (b) is histogram equalization, Figure 5 (c) is local histogram equalization, Figure 5 (d) is a homomorphic filter, Figure 5 (e) is a vein image obtained by using the vein image enhancement method of the present invention.

[0055] This paper proposes a vein image enhancement method based on multi-channel filters, and the parameters of the filter can be selected according to the width and direction of the vein network. A 4-way uniform symmetric filter with different center frequencies is used to develop vein information, and then the filtered image is fused into an integrated vein image using a reconstruction rule. Experimental results show that this method has a strong ability to improve low-quality vein images.

[0056] The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the attached claims.

Claims

1. A vein image enhancement method based on a multi-channel filter, characterized in that: The steps include: Step 1, obtaining a vein image; Step 2, the filter is, where k (=1, 2, 3, 4) is the channel index, θk (=kπ / 4) and f k denote the orientation and center frequency of the uniform symmetrical filter respectively; ρ = 1 / (2πσ x σ y ), θ is the direction of the filter, f0 is the center frequency of the filter, and σ x and σ y are the two axes of the elliptical Gaussian envelope, x θ and θ is the rotated version of the filter's coordinates (x, y); Step 3, obtain the filtered vein image rk(x, y) by convolving the input vein image I(x, y) with the uniform symmetric filter Gek(x, y) at the Kth channel, Where * indicates that the convolution is in two dimensions, and the directions of the uniform and symmetric filters are 0°, 45°, 90°, and 135° respectively; Step 4, image reconstruction, the four filtered images of the filter are at each point pi(x i ,y i ) to form a 4-dimensional array S i , assuming that R(x, y) represents the reconstructed image, the reconstruction rule is defined as: Where N is the total number of pixels of the vein image.

Images