A Scattering Medium Imaging Method Based on Orthogonal Linear Polarization Image Frequency Domain Processing and Polarization Correlation

Abstract

This invention discloses a scattering medium imaging method based on frequency domain processing and polarization correlation of orthogonal linearly polarized images, applied to image restoration in turbid water. The method first preprocesses the minimum intensity map in the orthogonal linearly polarized image using high-pass filtering in the frequency domain, followed by polarization preservation and correlation processing, ultimately restoring a clear image of objects in the scattering medium. This innovative method combines frequency domain processing and polarization correlation of orthogonal linearly polarized images for image enhancement in turbid water, effectively separating and removing low-frequency scattered light information, thus improving image contrast and clarity. This method is of great significance for improving imaging quality in scattering media.

Description

Technical Field

[0001] This invention relates to an image restoration method for scattering medium imaging in a scattering medium environment, and more particularly to a descattering imaging method based on polarization image frequency domain processing and polarization correlation, belonging to the field of scattering imaging. Background Technology

[0002] The scattering and absorption of light by particles in scattering media such as water and smog can significantly degrade image quality. In turbid water imaging environments, the most significant factor causing image quality degradation is the interference of backscattered light. Backscattered light is partially polarized, and its polarization characteristics differ from those of the target signal light. Therefore, based on the acquisition and processing of polarization information, the target signal light and backscattered light can be effectively separated, thereby suppressing the influence of backscattered light and ultimately achieving a significant improvement in image quality. For this reason, polarization imaging technology has become one of the effective methods for clear underwater imaging. The core of underwater polarization imaging technology is suppressing backscattered light. Existing underwater polarization imaging technologies mainly suppress backscattered light based on the difference in polarization information between backscattered light and target signal light in the spatial domain. However, the target signal light and backscattered light overlap in the spatial domain; relying solely on the polarization difference between the target signal light and backscattered light is insufficient to fully suppress backscattered light.

[0003] In linearly polarized images acquired by imaging detectors in turbid water, the backscattered light from the scattering medium with relatively high polarization is generally concentrated in the low-frequency components of the spectrum, which is the main reason for the degradation of image quality; while the target object information with relatively low polarization is mainly concentrated in the high-frequency components. Summary of the Invention

[0004] The purpose of this invention is to address the need for clear imaging in scattering medium environments. Utilizing the difference in the distribution of target object information and scattered light information across the image spectrum, a scattering medium imaging method based on orthogonal linear polarization image frequency domain processing and polarization correlation is proposed. This method separates the target light component from the backscattered light component of the scattering medium, and then further restores the target object image by combining it with a traditional polarization imaging physical model. Stepwise scattering suppression improves the image degradation problem during underwater imaging. In environments with high concentrations of scattering medium, it enhances the imaging quality of the target object, ultimately achieving clear imaging of objects in the scattering medium. This method is simple, universal, and highly effective.

[0005] Technical solution of the present invention

[0006] A scattering medium imaging method based on orthogonal linearly polarized image frequency domain processing and polarization correlation is applied to the clear restoration of degraded images in scattering media (such as turbid water). This method combines the frequency domain processing of orthogonal linearly polarized images with the polarization degree correlation between each pair of orthogonal linearly polarized images, applying it to the field of polarization image enhancement in scattering medium environments. Based on the polarization information of the target light and the scattered light and their distribution differences in the frequency domain, the target light component is separated from the backscattered light of the scattering medium, effectively separating and removing low-frequency scattered light information, obtaining more accurate light intensity information of the target object in the scattering medium, and improving image contrast and clarity. This method requires no background or prior knowledge, is robust, and has significant image information restoration effects. The specific implementation steps are as follows:

[0007] First, ensure the direction of the actively illuminated linearly polarized light remains unchanged. Adjust the direction of the front polarizer of the camera to be parallel and perpendicular to the direction of the actively illuminated linearly polarized light, respectively, to obtain two mutually orthogonal polarization images, namely the maximum intensity image I(x,y) and the minimum intensity image I(x,y). ⊥ (x,y), and calculate the polarization degree diagram P(x,y) of the two:

[0008]

[0009] 2. Minimum light intensity diagram I ⊥ (x,y) is Fourier transformed to the frequency domain and then preprocessed with a Gaussian high-pass filter to obtain the minimum intensity map after filtering. And the cutoff frequency of the filter used is determined by... Image detail enhancement evaluation function f EME The system automatically searches for the optimal solution to the objective function and calculates the linearly polarized image of the high-frequency portion of the minimum intensity map after filtering with the optimal cutoff frequency d0 filter.

[0010]

[0011]

[0012] Third, maintain the inter-polarization relationship between the images before and after filtering, and for the optimal cutoff frequency d0... The linear polarization image I of the high-frequency portion of the maximum intensity map is obtained through polarization degree correlation calculation. h (x,y):

[0013]

[0014] 4. Obtain the high-frequency component linear polarization image of the preprocessed maximum and minimum light intensity maps. h (x,y) and The maximum and minimum values ​​of the medium gray level are taken as the maximum value B of the backscattered light.h and minimum value And calculate the polarization degree P of the backscattered light. bs ;

[0015]

[0016]

[0017] 5. Based on the underwater polarization imaging model, from I h (x,y) and Calculate the total light intensity image I h (x,y) and polarization difference light intensity image ΔI h (x,y), from backscattered light information B h Estimate the intensity of backscattered light at infinity And calculate the transfer function t h (x,y), and finally the restored image L(x,y) is calculated.

[0018]

[0019]

[0020] Theoretical basis of this invention:

[0021] For the underwater polarization imaging model, the total light intensity I(x,y) received by the camera during underwater imaging can be expressed as:

[0022]

[0023] Wherein, the intensity of the light L(x,y) reflected from the target object, after being scattered and absorbed by a medium with a transfer function of t(x,y), reaches the detector as O(x,y); the backscattered light intensity B(x,y) is partially polarized light; and the backscattered light intensity at infinity underwater (i.e., at the background) is A. ∞ .

[0024] From equation (1), it can be derived that the actual reflected light L(x,y) of the target object, i.e., the clear underwater image, can be expressed as:

[0025]

[0026] Underwater polarization imaging technology is based on orthogonal polarization image pairs I(x,y) and I... ⊥ The light intensity signal I(x,y) received by the camera is:

[0027]

[0028] Image I(x,y) and I ⊥(x,y) is difficult to accurately represent the light intensity and polarization information of the original target object. Based on the spectral differences between the target reflected light and the backscattered light of the scattering medium in the polarization image, a frequency domain high-pass filter can be designed to preprocess the orthogonal polarization image.

[0029] Will I ⊥ (x,y) is transformed to the frequency domain using Fourier transform and then subjected to a Gaussian high-pass filter with a cutoff frequency of D0 to obtain the relatively concentrated high-frequency portion of the target signal light. Minimum intensity diagram after filtering. for:

[0030]

[0031] Where F is the Fourier transform and D(u,v) represents the distance between two points in the spectrum.

[0032] The polarization relationship of the image before preprocessing is as follows:

[0033]

[0034] Perform polarization correlation operations to maintain the polarization degree relationship between the images before and after preprocessing:

[0035]

[0036] Because the orthogonal polarization image after frequency domain filtering preprocessing affects I h (x,y) and It only contains the high-frequency components of the original image spectrum, so it can be combined with the traditional polarization scattering suppression model to better restore the image in the scattering medium. The steps are as follows.

[0037] The degree of polarization P of the backscattered light bs It can be represented as:

[0038]

[0039] To make P bs The maximum value of backscattered light, B, within the physically feasible region, can be calculated using the following formula. h and minimum value

[0040]

[0041] From the transfer function expression, it can be seen that when t h As (x,y) approaches 0, the following formula holds:

[0042]

[0043] For a low-biased object, the polarization difference parameter ΔO of the target light is 0. At this time, the transfer function t can be obtained by combining formula (1) and formula (7).h (x,y):

[0044]

[0045] Here, Δ represents the difference between polarization orthogonal images.

[0046] To achieve optimal image restoration, the optimal cutoff frequency d0 is automatically searched. The optimal cutoff frequency d0 of the filter should make... Image detail enhancement evaluation function f EME The value of (Enhancement MeasureEvaluation) is the largest, that is...

[0047]

[0048] The optimal cutoff frequency d0 corresponds to Correlate the high-frequency component of the minimum intensity map I h (x,y), that is

[0049]

[0050] Based on the underwater polarization imaging model and the corrected orthogonal polarization image, the final restored image L(x,y) of the target object is calculated:

[0051]

[0052] The beneficial effects and advantages of this invention are as follows:

[0053] 1. The method of the present invention combines frequency domain preprocessing of polarization images with polarization correlation of orthogonal linear polarization images, and is applied to the field of imaging of scattering media such as turbid underwater. It effectively separates and removes low-frequency scattered light information, suppresses the modulation of the actual reflected light of the object by the scattering medium, obtains more accurate object information, and improves the contrast and clarity of the restored image.

[0054] 2. The method of the present invention has a significant imaging effect in strong scattering environments, and the image preprocessing process makes up for the shortcomings of traditional polarization imaging methods in the poor imaging effect in strong scattering media.

[0055] 3. The method of the present invention is a fully automatic image processing method that does not require background or prior knowledge. It can automatically obtain the image restoration result in the final scattering medium, has strong robustness, and has a significant image information restoration effect. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of a polarization imaging system device in a scattering medium.

[0057] Figure 1The markings are explained as follows: 1-Light source, 2-First linear polarizer, 3-Scattering medium, 4-Target object, 5-Second linear polarizer, 6-Light intensity detector;

[0058] Figure 2 The flowchart of the frequency domain-based underwater polarization imaging algorithm;

[0059] Figure 3 This is the original light intensity image of the scattering medium;

[0060] Figure 4 (a) Linear polarization image of minimum intensity and (b) Linear polarization image of maximum intensity;

[0061] Figure 5 (a) is the filtered image of the minimum linear polarization intensity and (b) is the filtered image of the maximum linear polarization intensity.

[0062] Figure 6 (a) Image restoration effect of this method and (b) Image restoration effect of Schechner method;

[0063] Figure 7 (a) Original light intensity map, (b) Image restoration effect of this method, and (c) Image restoration effect of Schechner method in a scattering medium with higher turbidity. Detailed Implementation

[0064] Example

[0065] Appendix Figure 1 This is a diagram of an imaging system for scattering media (turbid water) involved in the method of this invention. The image restoration method for scattering media of this invention has been verified through experiments, and its specific algorithm flow is as follows: Figure 2 As shown.

[0066] The light emitted by the LED light source 1 is polarized by the first linear polarizer 2 to generate a beam of linearly polarized light as an active illumination source. The beam passes through the strong scattering medium 3 containing milk and water and irradiates the surface of the target object 4. After interacting with the target object and generating reflected light, it interacts with the scattering medium again and passes through the second linear polarizer 5 located in front of the light intensity detector 6 before being received by the detector.

[0067] The specific steps of the scattering medium imaging method based on orthogonal linear polarization image frequency domain processing and polarization correlation are as follows:

[0068] In the experiment, target object 4 was fixed inside a water tank, and water and milk were added to simulate a turbid scattering environment. The polarization direction of the first linear polarizer 2 in front of light source 1 was fixed. Two linearly polarized images, one of minimum light intensity and one of maximum light intensity, were obtained by rotating the first linear polarizer in front of the time-division imaging system. Figure 4 As shown;

[0069] Minimum intensity image I that highlights the target component ⊥ After Fourier transform, (x, y) undergoes high-pass filtering preprocessing according to Formula 4 to extract high-frequency components that concentrate the details of the target object. The minimum intensity map after filtering is then obtained through inverse Fourier transform. The minimum intensity maps before and after filtering are added together to obtain the result. And calculate the image quality evaluation function f EME The value of d0 is used to select the filter cutoff frequency corresponding to the maximum value according to Formula 12. After preprocessing with this filter as shown in Formula 13, and then performing an inverse Fourier transform, the high-frequency part of the minimum light intensity map corresponding to the optimal cutoff frequency is obtained. This achieves initial suppression of scattered light, such as... Figure 5 As shown in (a);

[0070] There is a polarization correlation between the maximum and minimum intensity maps as described in Formula 5. To avoid disrupting this relationship, the polarization degree P remains constant, and the high-frequency linear polarization map I of the maximum intensity map is calculated from the high-frequency linear polarization map of the minimum intensity map according to Formula 6. h (x,y), such as Figure 5 As shown in (b).

[0071] According to Formula 8, the maximum gray value of the high-frequency linearly polarized image of the maximum intensity map and the minimum gray value of the high-frequency linearly polarized image of the minimum intensity map are calculated respectively, and then combined with the traditional polarization imaging physical model to estimate the polarization degree P of the backscattered light. bs Backscattered light intensity at infinity and transfer function t h (x,y), further calculated according to Formula 14 to obtain the restored image of the target object, such as Figure 6 As shown in (a).

[0072] Image restoration in the scattering medium was performed using Schechner's traditional polarization imaging algorithm, and the final result is as follows: Figure 6 As shown in (b), the comparison shows that the image restored by this method has significantly improved contrast and sharpness, and the effect is significantly better than the restoration results of traditional methods.

[0073] Experiments were conducted in scattering media of different concentrations to verify the advantages of this method. Taking a high-concentration scattering medium as an example, the method was tested in light intensity where details were significantly degraded. Figure 7 In the restoration of (a), such as Figure 7 (c) Traditional methods have extremely limited restoration effects and the target information is difficult to distinguish, while this method effectively suppresses the modulation of the actual reflected light by the scattering medium, obtaining information such as... Figure 7 (b) shows clearer details of the target object.

Claims

1. A method for imaging scattering medium based on orthogonal linear polarization image frequency domain processing and polarization correlation, applied to clear recovery of degraded image in scattering medium; the method combines frequency domain information processing of orthogonal linear polarization image and polarization correlation between orthogonal linear polarization image pairs, and step by step suppresses modulation of scattering medium to actual reflection light of object; the method is applied to the field of polarization image enhancement in scattering medium environment, and specific implementation steps are as follows:

1. The first, to ensure the active illumination linearly polarized light direction, adjust the camera before the direction of the polarizer to be parallel, perpendicular to the active illumination linearly polarized light direction, respectively, to obtain two orthogonal polarized images, respectively, the maximum light intensity map And the minimum light intensity map , And the polarization degree map : ； The second minimum light intensity map The filtered minimum light intensity map obtained by the Gaussian high-pass filtering preprocessing And the optimal solution automatic search is performed on the cut-off frequency of the used filter to calculate the optimal cut-off frequency The linear polarization image corresponding to the high-frequency part of the minimum light intensity map ; 3. The optimal cutoff frequency corresponding to The polarization degree correlation calculation is performed to obtain a linear polarization image of the high-frequency part of the maximum light intensity image The polarization degree correlation calculation method is that the high-frequency part of the filtered minimum light intensity image corresponds to the high-frequency part of the maximum light intensity image The initial polarization degree relationship is still maintained, and the high-frequency part of the maximum light intensity image is obtained according to the unchanged polarization degree correlation : ； 4. Obtain high frequency component linear polarization image of the maximum and minimum intensity image after pre-processing and Maximum and minimum values of the middle gray scale as maximum and minimum values of backscattering light and minimum values and calculate the degree of polarization of backscattering light ; 5. From the underwater polarization imaging model, the total intensity image is calculated by and the polarization-difference intensity image is calculated by and the backscattering light intensity is estimated by the backscattering light intensity at infinity is estimated by and the transport function is calculated by and the final recovered image is calculated by .​ 2. The method of claim 1, wherein In step 2, the polarization image undergoes frequency domain filtering preprocessing. This is done by using the minimum intensity image... Transform to the frequency domain and perform Gaussian high-pass filtering to obtain the minimum intensity map after filtering. Evaluation function for image detail enhancement evaluate To improve image quality, the optimal cutoff frequency is determined by searching for the cutoff frequency used by the filter when the function value is maximized. Use the filter corresponding to the optimal cutoff frequency to obtain the minimum intensity map. Filtering is performed to obtain the high-frequency portion of the minimum intensity map. For subsequent restoration ； ； 。 3. The method of claim 1, wherein The maximum gray value corresponding to the maximum light intensity map in the fourth step is selected as the maximum value of the backscattering light The minimum gray value corresponding to the minimum light intensity map is selected as the minimum value of the backscattering light The polarization degree of the backscattering light is calculated : ； 。

Images