CN109671032B

CN109671032B - Method for constructing compensation factor equalization model according to image brightness characteristics

Info

Publication number: CN109671032B
Application number: CN201811546359.3A
Authority: CN
Inventors: 周国清; 刘小帆; 黄煜
Original assignee: Guilin University of Technology
Current assignee: Guilin University of Technology
Priority date: 2018-12-18
Filing date: 2018-12-18
Publication date: 2023-04-07
Anticipated expiration: 2038-12-18
Also published as: CN109671032A

Abstract

The invention discloses a method for constructing a compensation factor balance model according to image brightness characteristics, which comprises the steps of predicting the image brightness trend, establishing a brightness compensation model and extracting a compensation factor. In the image brightness trend prediction, row gray data with small image gray change is extracted, and the average gray value of each row in the effective row gray data is calculated; setting a corresponding brightness expected value in a brightness compensation model, carrying out quotient with the obtained average gray data to obtain a corresponding compensation value, and fitting the compensation data by using a fitting method to obtain a compensation model; in the process of extracting the compensation factors, the compensation factors with corresponding widths are extracted from the brightness compensation model according to the width of the image, so that the brightness balance is realized. The invention can effectively realize brightness balance according to the image characteristics, reduce the root mean square error of the image, shorten the effective gray level moving interval of the image and reduce the peak value of the radiation difference of the image.

Description

Method for constructing compensation factor equilibrium model according to image brightness characteristics

Technical Field

The invention relates to the field of image brightness equalization, in particular to the field of image brightness equalization realized by using a coefficient method, which is applied to remote sensing image brightness equalization processing.

Technical Field

With the development of image processing technology, the brightness equalization method has already become mature and perfect, and the method for processing the brightness difference of the image by the coefficient has developed rapidly and becomes the main research direction of the current brightness equalization processing. When the image has brightness difference, the original gray value of the image is changed under the action of the coefficient, the brightness balance of the image is realized, the method has the advantages of being simple and convenient to operate, free from the limitation of the size of the image and the number of the images and the like, and the method is widely applied to the fields of remote sensing image processing and the like.

The current methods for realizing brightness equalization by using coefficients include a method for adding correction coefficients and a traditional method for increasing brightness coefficients. The method for adding the correction coefficient processes the original gray value of the image in an addition mode by calculating the gray difference between the target area and the reference area and processing the gray difference, such as: the invention with patent number CN201510704720.0 discloses a method for calculating brightness correction coefficients corresponding to two sides of an image and correcting brightness values of pixels of the image, which has the defects that the calculation accuracy cannot be ensured, gray scale mutation is easily caused, and when an image is large in size, the calculation amount is increased accordingly. In the conventional method for increasing the brightness coefficient, a specific model is established according to certain parameters, and the original gray value of the image is processed in a multiplication mode, such as: the invention with patent number CN201710431459.0 discloses a method for performing luminance equalization processing on a filtered image by solving a specific formula to form a luminance coefficient, which has the defect that although the method is not affected by calculation errors, a fixed model cannot satisfy the diversity and self-particularity of the image. In order to satisfy the brightness characteristics of the image and improve the processing precision, it is necessary to improve and perfect the coefficient method.

Disclosure of Invention

The invention aims to provide a method for constructing a compensation factor equalization model according to image brightness characteristics, which can effectively adapt to the self-specificity of an image, reduce the calculated amount, ensure the image precision, avoid the generation of gray level jump and realize the optimization of image color equalization.

The content of the invention is as follows: a method for constructing a compensation factor balance model according to image brightness characteristics comprises an image brightness trend prediction model, a brightness compensation model and a compensation factor balance model. In the image brightness trend prediction model, performing row (or column) difference processing on an image according to image brightness characteristics, setting a threshold value, extracting row gray data with small image gray change, performing difference on the obtained row gray data again, eliminating data with large gray difference, calling the remaining gray data as effective row gray data, calculating the average gray value of each column of the effective row gray data, realizing the prediction of the image brightness change trend, and obtaining the image brightness trend prediction model; in the brightness compensation model, selecting a proper expected value according to the gray level activity interval of the prediction model, quoting the expected value and each row of average gray level values to obtain a compensation amount, and fitting the compensation amount data by using a fitting method to obtain an image brightness compensation model; in the compensation factor balance model, according to the width of the image frame of the image, the compensation factor with the corresponding width is extracted from the brightness compensation model to obtain a compensation factor balance model, and the compensation factor balance model is multiplied by the original gray value of the image to obtain a processed image, so that the brightness balance is realized.

The invention has the beneficial effects that: the method for constructing the compensation factor balance model according to the image brightness characteristics comprises an image brightness trend prediction model, a brightness compensation model and a compensation factor balance model. The image brightness trend prediction model comprises mathematical description of image brightness characteristics and effectively predicts the trend of image brightness change; the brightness compensation model comprises a function model established according to the brightness characteristics of the image, and the influence of the calculation precision on the image and the constraint of the fixed model on meeting the brightness characteristics of the image are eliminated; the compensation factor balance model has the advantages of adaptability of the method for adding the correction coefficient to the image and stability of the traditional brightness coefficient method.

Drawings

FIG. 1 is a diagram illustrating image luminance differences according to an embodiment of the present invention.

Fig. 2 is a flow chart of the present invention.

Fig. 3 is a flow chart of luminance trend prediction according to the present invention.

FIG. 4 is a diagram of a luminance prediction model according to an embodiment of the present invention.

FIG. 5 is a schematic diagram of a compensation model according to an embodiment of the present invention.

FIG. 6 is a diagram illustrating processing results according to an embodiment of the present invention.

Detailed Description

The following further describes the embodiments of the present invention with reference to the drawings. It is to be understood that the described embodiments of the invention are only some of the embodiments of the invention, and not all of the embodiments. Based on the embodiments of the present invention, researchers in this field do not make creative efforts, and all other embodiments belong to the protection scope of the present invention.

Example (b):

in this embodiment, images of the area of the island of Greenland are selected and equalized, and the images are captured by an ARGON satellite. The brightness of the image is not uniform due to the shooting environment and the artificial machinery. As shown in fig. 1, there is a distinct brightness band in the lower right corner of the image.

The specific steps of the luminance equalization provided by the present invention can refer to the flowchart (fig. 2):

step 1: reading the gray data of the image in MATLAB, and converting the gray data into a double format.

Step 2: the image is expressed by the formula Δ gray = gray (m) _i+1 ,n)-gray(m _i And n) carrying out difference processing line by line, wherein m represents the line number of the image, and n represents the column number of the image, and counting all the differential gray data.

And 3, step 3: and setting a threshold T according to actual conditions, and extracting the line number with the delta gray being less than T as effective line gray data.

And 4, step 4: establishing an image brightness prediction model by combining the graph 3, firstly counting gray values on the same column in effective row brightness data, carrying out differential processing again, eliminating gray data with larger deviation, and carrying out repeated experiments for a limited time to ensure modeling precision; then according to a weighted mean formula

Calculating the remaining gray data, where x represents the gray weighted average of the row, p represents the weight represented by the pixels in the row, g (m) represents the gray size of the pixels, and m represents the number of effective rows, and arranging the average gray data of each row according to the position to form a row number sequence, thereby finally obtaining the image brightness prediction model, as shown in fig. 4.

And 5: and setting an expected value P according to the gray level activity interval of the image brightness prediction model, and obtaining a compensation value of the image brightness prediction model by taking the quotient of the expected value and the gray level data in the image brightness prediction model.

Step 6: fitting each compensation value by using a fitting method to obtain a fitting curve f (x), wherein f (x) is a brightness compensation model of the image, and the obtained result is shown in fig. 5.

And 7: according to the image width n, extracting the number of compensation factors with the corresponding width n from an expression f (x) of a brightness compensation model, and multiplying the compensation factors by the original gray data of the image in a multiplication mode. If the image is processed by using the compensation factors, the effect is too bright or too dark, the number of the compensation factors is properly prolonged, and the compensation factors with the same width as the image width are selected to process the original gray value of the image according to the actual situation of the image brightness characteristics.

And step 8: and outputting a processing result, converting the data into a required type, and clearly finding that the brightness band is eliminated by the processed image as shown in fig. 6 after the obtained image is equalized.

Claims

1. A construction method of a compensation factor equalization model according to image brightness characteristics comprises the steps of image brightness trend prediction, brightness compensation establishment and compensation factor extraction, and is characterized by comprising the following concrete steps: in the image brightness trend prediction, an image brightness prediction model is established according to image brightness characteristics, and the change trend of the image gray value is predicted; in the establishment of a brightness compensation model, a compensation value is obtained by setting a quotient of an expected value and effective row gray data, and the brightness compensation model is established by using a fitting method; in the step of extracting the brightness factors, the image width is combined, the brightness compensation factors are extracted in a brightness compensation model, and the brightness compensation factors are integrated with the original gray value of the image; the image brightness trend prediction comprises the following steps: (1) Extracting the row gray data by using a differential technology to form effective row gray data; (2) Calculating the gray level average value of each column according to a weighted average formula; (3) Arranging the average gray value data of each row according to the position to form a row of number rows, and analyzing the image characteristics and the change trend; establishing the illumination compensation includes: (1) Setting a proper expected value according to the gray level activity interval of the image brightness prediction model; (2) Taking the quotient of the expected value and the gray data in the image brightness prediction model to obtain a compensation value of the image brightness prediction model; (3) Fitting each compensation value by using a fitting method to obtain a function expression of the brightness compensation model; extracting the compensation factor includes: (1) Extracting compensation factors of corresponding widths in a brightness compensation model expression according to the width of the image frame; (2) And under the condition of over-bright or over-dark processing effect, properly prolonging the quantity of the compensation factors, and intercepting and selecting the compensation factors which are consistent with the width of the image frame according to the actual condition of the image brightness characteristic to process the original gray value of the image.