CN109767399B - Underwater image enhancement method based on unsupervised color correction - Google Patents

Abstract

The invention belongs to the field of ocean engineering, and particularly relates to an underwater image enhancement method based on unsupervised color correction. The invention improves the limitation that the channel B is taken as a prominent channel, thereby effectively solving the problem of color distortion of the underwater image with non-blue color cast; the invention comprises the following steps: the system comprises an RGB color model, a contrast correction method, an HIS color model and an underwater image enhancement method based on unsupervised color correction. Compared with the common underwater image enhancement method, the method has the advantages that: the method is improved to be a self-adaptive method, and has good processing effect on different color cast; effectively balance the image, eliminate color cast, improve illumination and increase true color. The method is easy to implement, small in calculated amount and high in reliability, improves feasibility of underwater optical image recognition, and has positive significance for development of unmanned underwater vehicles in underwater operation tasks and the like in future.

Description

An underwater image enhancement method based on unsupervised color correction

Technical Field

The invention belongs to the field of marine engineering, and in particular relates to an underwater image enhancement method based on unsupervised color correction.

Background Art

Due to the complex and harsh underwater imaging environment, the underwater images captured generally have deficiencies such as low contrast, blur and color cast. The degradation of image visual quality will have a serious adverse impact on the performance of subsequent feature extraction and target recognition processes, so it is of great significance to enhance the preprocessing of underwater images.

In China, the Marine Technology Research Center of Harbin Engineering University obtained information such as the position and size of underwater targets in marine engineering experiments in 2003 and 2004 by processing optical acoustic images, and achieved great success, thus proving that the underwater image processing algorithm is authentic and reliable. The Optoelectronic Imaging Laboratory of Beijing Institute of Technology can obtain distance information by studying sequence images and reconstruct underwater three-dimensional images; in 2008, Huazhong University of Science and Technology successfully segmented underwater images using two-dimensional wavelet transform, achieving enhanced local directional contrast. In 2012, Chen Congping et al. proposed an effective low-contrast underwater image enhancement algorithm, which successfully improved high-frequency gain and suppressed low-frequency components.

In the deep sea, sunlight cannot be transmitted here. Here we can only rely on point light sources to capture images. The brightest light is in the center, and light is emitted from the surroundings. Even the naked eye can see it, but the farther away from the light source, the weaker the light. Therefore, when the light energy information is represented on the image, there will be an uneven grayscale distribution. Due to the influence of the complex underwater environment, light is constantly absorbed by the water body, resulting in a continuous decrease in light energy. The scattering of water molecules and impurities in the water causes the light to shift. It is precisely because of this that underwater imaging is so difficult and the information collected is incomplete. The scattering and absorption characteristics of underwater media in different wavelength bands are inconsistent. Because the water body reflects the blue of the sky, the underwater image basically appears blue, which leads to color distortion of the underwater image (green or blue).

Summary of the invention

The purpose of the present invention is to solve the deficiencies in the prior art and to provide an underwater image enhancement method based on unsupervised color correction, which can effectively overcome the problem of distortion when the original method processes non-blue cast images.

The beneficial effects of the present invention are:

The underwater image enhancement method based on unsupervised color correction of the present invention effectively identifies the color cast of the image, and solves the color distortion problem of the original method in processing underwater images with non-blue color cast. The present invention combines the RGB space with the HIS space, and adds the contrast correction method in each direction to effectively eliminate the color cast problem of the underwater image, and improve the illumination and true color of the original image. Compared with other methods, this method can significantly improve the visual quality of underwater images and consistently enhance the objective quality of underwater images.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a diagram showing the implementation steps of an unsupervised color correction method.

Figure 2(a) is a typical underwater target visual image of Figure A.

Figure 2(b) is a typical underwater target visual image in Figure B.

Figure 3(a) is the image of Figure A after being processed by the original method.

Figure 3(b) is the image of image B after being processed by the original method.

FIG. 4( a ) is an image of FIG. A after being processed by the present invention.

FIG4( b ) is an image of the B image after being processed by the present invention.

DETAILED DESCRIPTION

The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content described in the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.

With the exploration of ocean, lake and river resources, underwater images have become an important research field. However, due to the complex and harsh underwater imaging environment, the underwater images captured often have problems such as color distortion and low contrast. Underwater imaging can be expressed as a linear superposition of forward scattering components and backscattering components. Forward scattering causes blurring of image features, while backscattering obscures the details of the scene. Since each color sub-image has a different wavelength and energy level, it is absorbed at different rates in water. Light components with longer wavelengths, such as red or orange light, tend to be quickly absorbed in water, so underwater images usually appear green or blue tones. The degradation of image visual quality will have a serious adverse effect on the performance of subsequent feature extraction and target recognition processes. Image enhancement is a basic method for processing images and is one of the effective means to solve the problem of blurred and severely low quality underwater images. In order to ensure the integrity of image information, it is very necessary to develop an underwater image enhancement method with better processing effect.

In order to enhance the original underwater image, the present invention proposes an underwater image enhancement method based on unsupervised color correction, which includes the following main steps:

Step 1: Use the RGB color model to calculate and process the image. Let _IR (i, j), _IG (i, j), _IJ (i, j) be the red, green, and blue components of an RGB image of size M×N, respectively, where i＝1,…,M; j＝1,…,N. Then calculate the maximum and minimum pixel values of each color component R _max , R _min , G _max , G _min and B _max , B _min :

Use the above equation to get the most and least prominent color cast channels.

Step 2: Use a multiplier to determine the color channel to match the m-order to produce a balanced image. First, calculate the average value of each color component R _avg , G _avg and B _avg in the RGB color model:

Secondly, the prominent color cast channel is obtained by using step 1, and the high component is used to increase other colors to balance the image. When R _max is the largest, the image is red cast; when G _max is the largest, the image is green cast; when B _max is the largest, the image is blue cast. Based on the dominant color cast, two gain factors are calculated, the highest color channel is set to the target average, and a multiplier is used to determine the color channel to match the m-order to produce a balanced image. The proposed method uses two color channels to reduce the color cast of the affected image.

When the R channel is a prominent channel:

a＝R _avg /G _avg

b＝R _avg /B _avg

G'＝a×G

B'＝b×B

Where G and B are the pixel values in the original image, and G' and B' are the adjusted pixel values.

When the G channel is a prominent channel:

a＝G _avg /R _avg

b＝G _avg /B _avg

R'＝a×R

B'＝b×B

Where R and B are the pixel values in the original image, and R' and B' are the adjusted pixel values.

When channel B is the prominent channel:

a＝B _avg /R _avg

b＝B _avg /G _avg

R'＝a×R

G'＝b×G

Where R and G are the pixel values in the original image, and R' and G' are the adjusted pixel values.

Step 3: Apply contrast correction method to the processed image. The contrast correction formula is:

Where _P0 is the contrast-corrected pixel value; _Pi is the pixel value under consideration; a is the lower limit of 0; b is the upper limit of 255; c is the minimum pixel value currently in the image; d is the maximum pixel value currently in the image. The contrast correction method is applied to the upper side, the lower side, and the two sides in turn. When applied to the upper side, the lower limit is changed to the minimum value of the lowest color value component; when applied to the upper side, the upper limit is changed to the maximum value of the most prominent color channel; when applied to both sides, the formula remains unchanged.

Step 4: Convert the image from RGB space to HIS space. The conversion formula is:

Step 5: Apply contrast correction to the image in HIS color space. Here we only perform contrast correction on both sides, using the same method as step 3.

Through the operation of the above five steps, the enhanced underwater image is finally obtained. Compared with other existing technologies, the present invention effectively utilizes the high color cast in the image to add other colors to balance the image; the contrast correction method plays a vital role in color enhancement, because the use of this method can reduce the high color cast by stretching the high color cast color histogram to the minimum side. Similarly, the low color cast is increased by pulling the low color cast color histogram to the maximum side to generate high-quality images. The illumination and true color of the underwater image are increased by the intensity and saturation parameters of the HSI color model, so the image looks brighter and richer. The properties of the image are taken into account and the image is enhanced according to its characteristics rather than static standards, so the invention can better enhance the underwater image than some existing methods. Compared with the original method, the color distortion problem of the original method in processing underwater images with non-blue color cast is well solved.

Claims (6)

1. An underwater image enhancement method based on unsupervised color correction, characterized in that the method comprises the following steps: Step 1: Collect underwater images, and collect underwater target images under different background colors through the underwater image acquisition system; Step 2: Use the RGB color model to calculate the most and least prominent color cast channels of the image; Let _IR (i, j) be the red component of an RGB image of size M×N, let _IG (i, j) be the green component of an RGB image of size M×N, let _IJ (i, j) be the blue component of an RGB image of size M×N, where i = 1, …, M; j = 1, …, N; calculate the maximum and minimum pixel values of each color component, and the maximum value R _max of the red component is:

The minimum value R _min of the red component is:

The maximum value of the green component G _max is:

The minimum value of the green component G _min is:

The maximum value of the blue component B _max is:

The minimum value B _min of the blue component is:

Get the most and least prominent color cast channels; Step 3: Use a multiplier to determine the color channels to match the m-order to produce a balanced image; Calculate the average value of the red component, blue component and green component in the RGB color model; The average value of the red component R _avg in the RGB color model is expressed as:

The average value of the blue component G _avg in the RGB color model:

The average value of the green component B _avg in the RGB color model:

Utilize the color cast channel and use the high component to increase other colors to balance the image. When R _max is maximum, the image is red cast; when G _max is maximum, the image is green cast; when B _max is maximum, the image is blue cast; based on the dominant color cast, calculate the first gain numerator and the second gain numerator, set the highest color channel to the target average, and use a multiplier to determine the color channel to match the m-order to produce a balanced image; When the R channel is a prominent channel, the first gain molecule a _R of the R channel and the second gain molecule b _R of the R channel are: a _R = R _avg / G _avg b _R = R _avg / B _avg G'＝ _aR ×G B'＝ _bR ×B Where G and B are the pixel values in the original image, and G' and B' are the adjusted pixel values; When the G channel is a prominent channel, the first gain molecule a _G of the G channel and the second gain molecule b _G of the G channel are: a _G = G _avg / R _avg b _G = G _avg / B _avg R'＝ _aG ×R B'＝ _bG ×B Where R and B are the pixel values in the original image, and R' and B' are the adjusted pixel values; When channel B is a prominent channel, the first gain molecule a _B of channel B and the second gain molecule b _B of channel B are: a _B = B _avg / R _avg b _B = B _avg / G _avg R'＝a _B ×R G'＝b _B ×G Where R and G are the pixel values in the original image, and R' and G' are the adjusted pixel values; Step 4: Apply the contrast correction method to the processed image; Step 5: Convert the image from RGB space to HIS space; Step 6: Apply contrast correction method to the image in HIS color space. 2. According to claim 1, an underwater image enhancement method based on unsupervised color correction is characterized in that the underwater image acquisition system in step 1 includes an underwater camera, underwater lighting equipment, an image acquisition card, a desktop computer, an independent 12V DC power supply and an image processing platform; the underwater camera and the underwater lighting equipment are both placed underwater, the 12V DC voltage is placed on the shore to provide power for the underwater camera and the underwater lighting equipment at the same time, and the image acquisition card is installed on the PCI-E slot of the desktop computer. 3. The underwater image enhancement method based on unsupervised color correction according to claim 1, characterized in that the contrast correction method in step 4 comprises the following steps: Step 4.1: Contrast correction on the upper side; Step 4.2: Contrast correction on the lower side; Step 4.3: Contrast correction on both sides. 4. The underwater image enhancement method based on unsupervised color correction according to claim 3, characterized in that the contrast correction on the upper side in step 4.1 is expressed as the following formula:

The contrast correction for the lower side described in step 4.2 is expressed as follows:

The contrast correction on both sides described in step 4.3 is expressed as:

Where _P0 is the contrast-corrected pixel value; _Pi is the pixel value under consideration; a is the lower limit of 0; b is the upper limit of 255; c is the minimum pixel value currently present in the image; d is the maximum pixel value currently present in the image. 5. The underwater image enhancement method based on unsupervised color correction according to claim 4 is characterized in that, in step 5, the image is converted from RGB space to HIS space, and the conversion formula is:

6. The underwater image enhancement method based on unsupervised color correction according to claim 5 is characterized in that the contrast correction method in step 6 is the same as the contrast correction method in step 4, but the contrast correction is only performed on two sides.

Images