A photometry method based on information entropy

Abstract

The present invention discloses a photometry method based on information entropy, comprising the steps of: (1) preprocessing an original image; (2) calculating the probability distribution and cumulative probability distribution of a certain pixel value t on the image processed by step (1); (3) calculating the entropy of the pixel value t, and calculating the pixel threshold for extracting effective information on the image based on the entropy; (4) calculating the photometry value of the image based on the pixel threshold for extracting effective information on the image obtained by step (3) combined with the position weight of a certain pixel point on the image and the value range weight of its pixel value in the pixel value range on the image. The present invention performs Gaussian low-pass filtering on the image, and calculates the entropy of a certain pixel value based on the entropy, and determines the pixel threshold for extracting effective information on the image based on the calculated entropy, thereby obtaining truly effective information on the image, and furthermore, the calculation of the photometry value is more accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of image detection, and in particular to a light measurement method based on information entropy. Background Art

[0002] In C-arm imaging, the dose rate radiated to the flat-panel detector is one of the important factors affecting the imaging quality. In the actual imaging process, the dose rate is mainly restricted by the tube voltage and tube current. Usually, an automatic dimming algorithm is used to adjust the tube voltage and tube current in real time according to the photometric value of the previous frame or several previous frames. Therefore, in a fixed test scene, the photometric value ultimately determines the size of the dose rate.

[0003] Traditional photometry methods often consider the average pixel value of the entire frame as the final photometry value, but the photographed object cannot always occupy the entire frame; secondly, when X-rays penetrate an object, the dose rate becomes lower because the radiation is absorbed, so generally lower pixel value information represents truly effective information. In addition, considering that the human eye pays more attention to the center of the image, the photometry value obtained without considering the pixel position cannot represent the true perception of the human eye. Summary of the invention

[0004] Purpose of the invention: In view of the above-mentioned shortcomings, the present invention proposes a photometry method based on information entropy, which solves the problems that in X-ray imaging, the traditional photometry method does not take into account that the effective photometry area often does not occupy the entire frame image, the effective information of the image is carried in the pixel information with a relatively small value range, and the human eye is more sensitive to the central area, which causes the photometry value to be large and the actual image does not conform to the human eye's perception.

[0005] Technical solution:

[0006] A photometry method based on information entropy, comprising the steps of:

[0007] (1) Preprocess the original image;

[0008] (2) Calculate the probability distribution and cumulative probability distribution of a pixel value t on the image processed by step (1);

[0009] (3) Calculate the entropy of the pixel value t, and calculate the pixel threshold for extracting effective information on the image based on it;

[0010] (4) The pixel threshold is extracted based on the effective information on the image obtained in step (3), combined with the position weight of a certain pixel point on the image and the range weight of its pixel value in the pixel value range on the image, to calculate the light metering value of the image.

[0011] The step (1) specifically comprises: performing fast Fourier transform on the original image to obtain a frequency domain image, performing Gaussian low-pass filtering on the obtained frequency domain image, and performing inverse Fourier transform on the obtained frequency domain image to obtain a processed time domain image.

[0012] The filter kernel of the Gaussian low-pass filter is:

[0013]

[0014] Among them, (m,n) represents the coordinates of the image boundary pixel points, D(m,n) represents the distance from the image boundary pixel points to the image center point, and D0 represents the cutoff frequency.

[0015] In step (2), the probability distribution p of a pixel value t on the image processed by step (1) is calculated. t And the cumulative probability distribution P t The details are as follows:

[0016]

[0017] Among them, 0≤t≤max, max represents the maximum pixel value obtained in the image imaged on the detector; card({z ij =t}) represents the pixel value z on the image ij is the number of pixels of t, max x 、max y Respectively represent the number of pixels in the rows and columns of the image.

[0018] In step (3), the entropy of pixel value t is calculated Specifically:

[0019]

[0020] Among them, p m Represents the probability distribution of pixel value m on the image;

[0021] Then the pixel threshold for extracting effective information on the image is

[0022] In step (3), a pixel point (x i ,y j ) The position weight u on the image ij Specifically:

[0023]

[0024] Among them, U1<U2<U3, R1 and R2 represent the parameters that need to be calibrated respectively; take the center point of the image as the center, and construct two concentric circles with R1 and R2 as the radius to form an image area with three position weights;

[0025] Then the pixel point (x i ,y j ) in the pixel value range on the image ij Specifically:

[0026]

[0027] Among them, V1<V2<V3, a1 and a2 are set according to the requirements, and 0 <a1<a2<max。

[0028] The weights in the image region of the three weights are set to U1=1, U2=2, and U3=3.

[0029] a1 is set to a2 is set to And the pixel value z of pixel point (i, j) ij The range weight v of the pixel value range on the image ij Assume that V1=1, V2=2, and V3=3.

[0030] Said Said

[0031] In step (3), the photometric value P of the image is calculated as follows:

[0032]

[0033] Among them, z ij ≤T.

[0034] Beneficial effect: The present invention performs Gaussian low-pass filtering on the image and calculates the entropy of a certain pixel value based on the calculated entropy, and determines the pixel threshold for extracting effective information on the image based on the calculated entropy, thereby obtaining truly effective information on the image, and furthermore, the calculation of the light metering value is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a flow chart of the present invention.

[0036] Figure 2 A schematic diagram of an image coordinate system established for the processed time domain image obtained by the present invention;

[0037] Figure 3 It is a schematic diagram of position weight of the present invention;

[0038] Figure 4 Schematic diagram of the range weight of the present invention. DETAILED DESCRIPTION

[0039] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments.

[0040] Figure 1 It is a flow chart of the present invention, as Figure 1 As shown, the photometry method based on information entropy of the present invention comprises the following steps:

[0041] (1) Import the original image img and perform fast Fourier transform on the original image img to obtain the frequency domain image Prepare for low-pass filtering of images;

[0042] (2) To prevent the ringing after filtering from interfering with the light measurement, select the frequency domain image img f Perform Gaussian low-pass filtering and perform inverse Fourier transform after filtering to convert the image to the time domain to obtain the processed time domain image img t ′;

[0043]

[0044] img f ′=img f *H(m,n)

[0045] Among them, img f ′ represents the frequency domain image img f The image obtained by Gaussian low-pass filtering, H(m,n) represents the filter kernel, which is expressed as follows:

[0046]

[0047] Where (m,n) represents the coordinates of the image boundary pixel points, D(m,n) represents the distance from the image boundary pixel points to the image center point, and D0 represents the cutoff frequency;

[0048] (3) Figure 2 Schematic diagram of the image coordinate system established for the processed time domain image of the present invention, a two-dimensional coordinate system is established with the center point of the processed time domain image as the coordinate origin. i ,y j ,z ij ,u ij ,v ij ) represents the pixel (x i ,y j ) is the pixel value z ij , the position weight is u ij , the range weight is v ij ;

[0049] Then calculate the time domain image img obtained in step (2) t ′Pixel value z ij is the probability distribution p of t t And the cumulative probability distribution Pt , where t∈img t ′, 0≤t≤max, max represents the maximum pixel value obtained in the image imaged on the detector;

[0050]

[0051] Among them, card({z ij =t}) represents the pixel value z ij = the number of pixels of t, max x 、max y Respectively represent the time domain image img t ′The number of pixels in the row and column;

[0052] (4) Calculate the entropy of pixel value t And the pixel threshold T for extracting effective information on the image is obtained based on entropy calculation;

[0053]

[0054] Among them, p m Represents the time domain image img t The probability distribution of the pixel value m on ′;

[0055] (5) Calculate the pixel point (x i ,y i ) The position weight u on the image ij and its pixel value z ij The range weight v in the pixel value range on the image ij ,like Figure 3 , 4 As shown, the metering value P of the image is calculated based on the final weight:

[0056]

[0057] Among them, R1 and R2 represent the parameters that need to be calibrated respectively; take the center point of the image as the center, and construct two concentric circles with R1 and R2 as the radius, so as to form the image area with three position weights; a1 and a2 are set according to the requirements, and 0 <a1<a2<max;

[0058] In the present invention,

[0059] In the present invention, the weights in the image region of the three weights only need to satisfy U1<U2<U3, and can be set to U1=1, U2=2, and U3=3.

[0060] In the present invention, a1 can be set to a2 can be set to

[0061] In the present invention, the pixel point (x i ,y j )’s pixel value z ij The range weight v in the pixel value range on the image ij It only needs to satisfy V1<V2<V3, and can be set to V1=1, V2=2, V3=3.

[0062] The present invention can obtain truly effective information on the image and calculate the light measurement value more accurately.

[0063] The preferred embodiments of the present invention are described in detail above, but the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations (such as quantity, shape, position, etc.) can be made to the technical scheme of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.

Claims

1. A photometry method based on information entropy, characterized in that: Includes steps: (1) Preprocess the original image; (2) Calculate the probability distribution p of a pixel value t on the image processed by step (1) t And the cumulative probability distribution P t , as follows: Where, 0≤t≤max, max represents the maximum pixel value obtained in the image after processing in step (1); card({z ij =t}) represents the pixel value z on the image ij is the number of pixels of t, max x 、max y Respectively represent the number of pixels in the rows and columns of the image; (3) Calculate the entropy of pixel value t Specifically: Among them, p m Represents the probability distribution of pixel value m on the image; Then the pixel threshold for extracting effective information on the image is Based on this, the pixel threshold for extracting effective information on the image is calculated; (4) Extract pixel threshold based on the effective information on the image obtained in step (3) and combine it with a certain pixel point (x i ,y j ) The position weight u on the image ij and its pixel value in the pixel value range on the image. ij , specifically: Among them, U1<U2<U3, R1 and R2 represent the parameters that need to be calibrated respectively; take the center point of the image as the center, and construct two concentric circles with R1 and R2 as the radius to form an image area with three position weights; Among them, V1<V2<V3, a1 and a2 are set according to the requirements, and 0 <a1<a2<max; Based on this, the image's metering value P is calculated as follows: Among them, z ij ≤T.

2. The photometry method based on information entropy according to claim 1, characterized in that: The step (1) specifically comprises: performing fast Fourier transform on the original image to obtain a frequency domain image, performing Gaussian low-pass filtering on the obtained frequency domain image, and performing inverse Fourier transform on the obtained frequency domain image to obtain a processed time domain image.

3. The photometry method based on information entropy according to claim 2, characterized in that: The filter kernel of the Gaussian low-pass filter is: Among them, (m,n) represents the coordinates of the image boundary pixel points, D(m,n) represents the distance from the image boundary pixel points to the image center point, and D0 represents the cutoff frequency.

4. The photometry method based on information entropy according to claim 1, characterized in that: The weights in the image area of ​​the three position weights are set to U1=1, U2=2, and U3=3.

5. The photometry method based on information entropy according to claim 1, characterized in that: a1 is set to a2 is set to And the pixel (x i ,y j )’s pixel value z ij The range weight v of the pixel value range on the image ij Assume that V1=1, V2=2, and V3=3.

6. The photometry method based on information entropy according to claim 1, characterized in that: Said Said

Images