A brightness correction method for underwater active optical images based on relative radiometric correction

Abstract

The present invention discloses an underwater active optical image brightness correction method based on relative radiation correction. The method utilizes fitting parameters to compensate for underwater image channels, performs channel-by-channel brightness uniformity correction on the original image, eliminates image radiation distortion caused by optical path difference and light source non-uniformity, and restores relatively consistent illumination of images in each channel. The method has good adaptability and practicability, and has obvious restoration effect, and can be widely used in the restoration of underwater images. In the transformation process, relevant transformation parameters are automatically calculated according to the characteristics of the image itself, thereby ensuring the adaptability of the method of the present invention to different images.

Description

Technical Field

[0001] The present invention relates to the technical field of image restoration, and in particular to an underwater active optical image brightness correction method based on relative radiation correction. Background Art

[0002] The penetration depth of natural light with the sun as the light source in seawater is tens of meters in oceanic Class I water bodies and only within a few meters in nearshore Class II water bodies. Therefore, when detecting the ocean floor at a depth of more than tens of meters, especially more than 100 meters, the sunlight is very weak or even disappears, and only active light sources can be used for illumination. Since point light sources have great advantages in manufacturing, assembly, use and maintenance, point light sources are often used as illumination sources for underwater active optical imaging detection. However, when the illumination light contacts the seabed and reflects from the ground objects and propagates back to the camera sensor, the background light intensity of the entire imaging area changes unevenly, and decays around the strongest point of the light source. This causes dark corners to appear around the image, resulting in uneven overall brightness of the captured image, and there are a large number of such images in underwater detection tasks. Therefore, the study of underwater image quality improvement algorithms helps to optimize the visual effects of underwater images and improve underwater observation efficiency, which is of great significance to underwater exploration and subsequent applications.

[0003] At present, many methods have been developed, such as histogram transformation, frequency domain operation, and Retinex theory. The histogram transformation method has the advantages of simple algorithm and high efficiency. By setting the histogram grayscale transformation function, the histogram has a specific distribution, thereby improving the image quality. Although the visual quality of underwater images has been improved, obvious artifact areas often appear due to excessive enhancement in some areas, which amplifies the noise in the image. The frequency domain operation performs Fourier transformation on the image, converts the real domain into the frequency domain for processing, and then converts it back to the spatial domain to obtain the output. The transform domain method can effectively enhance the high-frequency components of the image and thus improve the quality of underwater images. However, it is difficult to accurately measure optical parameters in the imaging environment. The quality of underwater images is mainly affected by the scattering effect, absorption and noise of light. The transform domain method has certain limitations in solving the problem of image degradation. The Retinex theory is used to separate the illumination component and the reflection component from the observed image, or the relationship between the illumination component and the reflection component is changed by eliminating the influence of illumination according to specific processing goals and requirements, thereby specifically enhancing the image. Although the Retinex theory can improve the overall brightness and contrast of the image to a certain extent, the restored images often have problems such as halo and unnaturalness. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide an underwater active optical image brightness correction method based on relative radiation correction, which can effectively eliminate the radiation distortion caused by factors such as uneven light source and different optical path in each channel, so that the center and edge of the image have relatively consistent brightness.

[0005] In order to solve the above technical problems, the present invention provides an underwater active optical image brightness correction method based on relative radiation correction, comprising the following steps:

[0006] Step 1: Separate the input image f(x, y) into three channels f R (x, y), f G (x, y), f B (x, y);

[0007] Step 2: Use polynomials to calculate the image f of each channel n (x, y), the brightness of n∈{R, G, B} is respectively subjected to surface fitting;

[0008] Step 3: Use η n (x, y) represents the brightness fitting surface of the reference image; select the fitting surface η n The attenuation of the vertex (x, y) (i.e. the point with the maximum brightness) is 0, and single-point correction is performed using this point as the reference point;

[0009] Step 4: After obtaining the brightness compensation amount of each pixel, directly add the compensation amount to the original image according to the pixel position, and output the brightness compensated image g n (x, y), calculated as follows:

[0010] g n (x, y) = f n (x,y)+δ n (x, y);

[0011] Step 5: Obtain the images after brightness correction of each channel and synthesize them into an RGB three-channel image g(x, y) again.

[0012] Preferably, in step 2, the polynomial surface fitting calculation formula is as follows:

[0013]

[0014] The setting q can specify the degree of the polynomial, a ij is the coefficient of a general polynomial, and (x, y) represents the pixel position in the image.

[0015] Preferably, in step 3, using n n (x, y) represents the brightness fitting surface of the reference image; select the fitting surface η nThe attenuation of the vertex (i.e., the point with the maximum brightness) of (x, y) is 0, and the single-point correction is performed with this point as the reference point. Specifically, let the maximum brightness value of the fitting surface be I Max , assuming that the plane γ passing through the vertex of the surface n (x, y) is perpendicular to the camera optical axis, then the brightness compensation amount δ for each pixel in the image n (x, y) is defined as the value η of the pixel on the fitting surface n (x, y) and in the plane γ n The difference between (x, y) is calculated as follows:

[0016] δ n (x, y) = γ n (x, y)-η n (x, y).

[0017] Preferably, for a grayscale image, step 1 is skipped and steps 2 to 5 are directly performed on a single channel.

[0018] The beneficial effects of the present invention are as follows: (1) the fitting parameters are used to compensate the underwater image channels, the brightness uniformity of the original image is corrected for each channel, the image radiation distortion caused by the optical path difference and the non-uniformity of the light source is eliminated, and the relatively consistent illumination of the images of each channel is restored. The method has good adaptability and practicality, and the restoration effect is obvious, and can be widely used in the restoration of underwater images; (2) in the transformation process, the relevant transformation parameters are automatically calculated according to the characteristics of the image itself, thereby ensuring the adaptability of the method of the present invention to different images. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the method flow of the present invention.

[0020] Figure 2 This is an underwater image of the present invention.

[0021] Figure 3 The figure is a result diagram of using the method of the present invention. DETAILED DESCRIPTION

[0022] like Figure 1 As shown, a method for underwater active optical image brightness correction based on relative radiation correction includes the following steps:

[0023] Step 1: Separate the input image f(x, y) into three channels f R (x, y), f G (x, y), f B (x, y);

[0024] Step 2: Use polynomials to calculate the image f of each channel n(x, y), n∈{R, G, B} brightness are respectively fitted with a surface, and the polynomial surface fitting calculation formula is as follows:

[0025]

[0026] The setting q can specify the degree of the polynomial, a ij is the coefficient term of the general polynomial, (x, y) represents the pixel position in the image, i and j are the polynomial indices respectively;

[0027] Step 3: Use η n (x, y) represents the brightness fitting surface of the reference image; select the fitting surface η n The attenuation of the vertex (i.e., the point with the maximum brightness) of (x, y) is 0, and single-point correction is performed with this point as the reference point. The maximum brightness value of the fitting surface is set to I Max , assuming that the plane γ passing through the vertex of the surface n (x, y) is perpendicular to the camera optical axis, then the brightness compensation amount δ for each pixel in the image n (x, y) is defined as the value η of the pixel on the fitting surface n (x, y) and in the plane γ n The difference between (x, y) is calculated as follows:

[0028] δ n (x, y) = γ n (x, y)-η n (x, y);

[0029] Step 4: After obtaining the brightness compensation amount of each pixel, directly add the compensation amount to the original image according to the pixel position, and output the brightness compensated image g n (x, y), calculated as follows:

[0030] g n (x, y) = f n (x,y)+δ n (x, y);

[0031] Step 5: Get the image after brightness correction of each channel and synthesize it into RGB three-channel image g(x, y) again.

[0032] In step 2, a polynomial is used to fit the attenuation surface; the present invention uses a polynomial to perform surface fitting on the brightness of each channel image. In order to explore the appropriate degree of the polynomial, we conducted a series of experiments, using quadratic, cubic, quartic and quintic polynomials to perform surface fitting on the brightness of each channel in the baseline reference image. Through experiments, we found that the fitting effect is best when a quintic polynomial is used. However, due to the high computational complexity of high-order polynomials, the fitting results of the quartic polynomial and the quintic polynomial are very similar, and the quartic polynomial can obtain good fitting effects in many different experiments, so the quartic polynomial is used for surface fitting. The polynomial fitting formula is as follows:

[0033]

[0034] The setting q can specify the degree of the polynomial, a ij is the coefficient of a general polynomial, and (x, y) represents the pixel position in the image.

[0035] In step 3, the brightness correction is performed using the fitted surface; η is used to represent the brightness fitting surface of the reference image, which describes the relative distribution of attenuation caused by various factors at different positions of the underwater optical image. Since the underlying surface of the reference image is a uniform surface formed by the same material, when the incident light intensity is the same, its image should have the same brightness. In order to obtain the total attenuation of the seawater to the downward and upward light, we select the vertex of the fitting surface (i.e., the point with the maximum brightness) as 0, and use this point as the reference point for single-point correction. Let the maximum brightness value of the fitting surface be I Max , assuming that the plane γ passing through the vertex of the surface is perpendicular to the camera optical axis, then the brightness compensation amount δ(x, y) of each pixel on the image is defined as the difference between the value η(x, y) of the pixel on the fitting surface and the value on the plane γ(x, y), that is:

[0036] δ(x,y)=γ(x,y)-η(x,y)

[0037] After obtaining the brightness compensation amount of each pixel, the compensation amount can be directly added to the original image according to the pixel position to obtain the brightness compensation result.

[0038] Figure 2 Shown is a reference image, Figure 3 The brightness correction results of the same batch of images are compared with the original images. It can be found that the brightness of the compensated images is uniform overall, eliminating the brightness attenuation caused by the different total optical paths. In particular, the brightness of the four corners far from the illumination center of the original image is well compensated, and the image details are highlighted. It is obvious that the effect after processing using the algorithm disclosed in the present invention is the best. It provides an effective method for automatic restoration of underwater image quality.

Claims

1. A method for underwater active optical image brightness correction based on relative radiation correction, characterized in that: The steps include: Step 1: Separate the input image f(x, y) into three channels f R (x, y), f G (x, y), f B (x, y); Step 2: Use polynomials to calculate the image f of each channel n (x, y), the brightness of n∈{R, G, B} is respectively subjected to surface fitting; Step 3: Use η n (x, y) represents the brightness fitting surface of the reference image; select the fitting surface η n The vertex of (x, y), i.e. the point with the maximum brightness, has an attenuation of 0, and this point is used as the reference point for single-point correction; n (x, y) represents the brightness fitting surface of the reference image; select the fitting surface η n The attenuation of the vertex (x, y) is 0, and the single-point correction is performed with this point as the reference point: Set the maximum brightness value of the fitting surface to be I Max , assuming that the plane γ passing through the vertex of the surface n (x, y) is perpendicular to the camera optical axis, then the brightness compensation amount δ for each pixel in the image n (x, y) is defined as the value η of the pixel on the fitting surface n (x, y) and in the plane γ n The difference between (x, y) is calculated as follows: d n (x, y)=γ n (x,y)-n n (x,y); Step 4: After obtaining the brightness compensation amount of each pixel, directly add the compensation amount to the original image according to the pixel position, and output the brightness compensated image g n (x, y), calculated as follows: g n (x,y)=f n (x,y)+δ n (x, y)? Step 5: Get the image after brightness correction of each channel and synthesize it into RGB three-channel image g(x, y) again.

2. The underwater active optical image brightness correction method based on relative radiation correction according to claim 1, characterized in that: In step 2, the polynomial surface fitting calculation formula is as follows: Where q is set to the degree of the specified polynomial, a ij is the coefficient of a general polynomial, and (x, y) represents the pixel position in the image.

3. The underwater active optical image brightness correction method based on relative radiation correction as claimed in claim 1, characterized in that: For grayscale images, skip step 1 and directly implement steps 2 to 5 on a single channel.

Images