A fast on-board dim light image enhancement method

Abstract

The application provides a fast airborne dark light image enhancement method, comprising the following steps: 1, transforming the input airborne dark light image to HSV color space through color conversion, and implementing dark light enhancement algorithm on the V channel component of the HSV color space; 2, filtering and processing the brightness component through a boundary-preserving smoothing operator; 3, performing pre-enhancement processing on the V channel component through the boundary-preserving smoothing component obtained in step 2; 4, enhancing the contrast of the image obtained in step 3 through the high-order fractional model enhancement proposed in the application; and 5, performing brightness enhancement on the result obtained in step 4 through the adaptive gamma transformation proposed in the application. The method proposed in the application not only can obtain excellent enhancement performance, but also is very efficient in processing speed; the dark part enhancement performance is excellent, and meets the requirements of dark part target recognition, dark part information collection and acquisition under the airborne scene.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of airborne digital image processing, in particular to a fast airborne dark light image enhancement method. BACKGROUND

[0002] When a high-speed or supersonic aircraft is flying, in order to ensure that the images of the ground or space scenes collected do not appear motion blur, it is usually necessary to shorten the single image signal acquisition time of the airborne camera. For the same aperture and sensor, the image frame rate collected per unit time increases with the increase of single image acquisition time, and the shorter the single image acquisition time, the weaker the response of the sensor to the light intensity, especially in dark light environment, the airborne image will appear insufficient brightness, which causes the weakening of the target features in the collected image, which is not conducive to the later target detection, observation, identification and classification, etc., so it is necessary to enhance the airborne dark light image, but the calculation efficiency and performance of the image enhancement method need to be considered at the same time.

[0003] At present, the traditional dark light enhancement methods include gamma transformation method, histogram equalization method, wavelet transformation method, etc., but the above methods are difficult to obtain excellent calculation efficiency and performance at the same time. Researchers have proposed Retinex enhancement theory by studying the color constancy characteristics of human visual system, and a large number of model-based and learning-based Retinex enhancement methods have been produced, but Retinex enhancement method usually needs high calculation amount to ensure its excellent calculation performance. SUMMARY

[0004] In order to solve the shortcomings that the calculation efficiency and performance of the traditional methods such as gamma transformation method, histogram equalization method and wavelet transformation method cannot be compatible, the present application designs a fast airborne dark light image enhancement method, which can quickly enhance the airborne dark light image and ensure its calculation efficiency and performance.

[0005] The technical scheme for achieving the purpose of the application is as follows: a fast airborne dark light image enhancement method, comprising the following steps:

[0006] S1, inputting a dark light image, performing color conversion from RGB color space to HSV color space;

[0007] S2, extracting the V channel component of the dark light image after color conversion to obtain a single channel image ;

[0008] S5, based on a high-order fractional model, enhancing the contrast of the single channel image to obtain a contrast-enhanced dark light image ;

[0009] S6, enhancing the contrast of the dark-light image based on an adaptive gamma transformation method to obtain a dark-light enhanced image.

[0010] In one embodiment, before enhancing the contrast of the single-channel image , the method further comprises:

[0011] S3, filtering the single-channel image based on an edge-preserving smoothing operator to obtain an edge-preserving smoothing image wherein the edge-preserving smoothing operator is represented by

[0012] S4, since the high-frequency information of the edge of the image collected in the dark-light environment is weak, the edge-preserving smoothing image is pre-enhanced to obtain a pre-enhanced image , and the pre-enhanced image is wherein the high-frequency information pre-enhancement coefficient is represented by

[0013] Further, in the step S5, the high-order fractional model is wherein the dark-light contrast enhanced image is represented by , and n is a positive integer greater than 1.

[0014] Further, in the step S6, the dark-light enhanced image is wherein the adaptive gamma coefficient is represented by

[0015] The present application compares the contrast enhancement effect of the single-channel image by the existing Retinex model and the high-order fractional model in the present embodiment:

[0016] The Retinex theory considers that an image can be decomposed into a reflection component independent of illumination and a brightness component related to the intensity of ambient light. First, the Retinex decomposition model is: wherein the pre-enhanced image is represented by , the reflection component is represented by , and the brightness component is represented by The image enhanced by the Retinex model is: wherein the traditional gamma transformation coefficient is represented by

[0017] The edge-preserving smoothing image with insufficient high-frequency information is regarded as the brightness component , i.e. , and the following equation is obtained:​​​​​​​​ Reflection component: .

[0018] Through the above analysis, with the help of the Retinex decomposition model, the dark light enhanced image in the specific embodiment can be changed to: .

[0019] From the above formula, it can be proved that when The fast airborne dark light image enhancement method proposed in the application is equivalent to the Retinex model method.

[0020] Compared with the prior art, the beneficial effects of the application are that the fast airborne dark light image enhancement method designed in the application not only can obtain excellent enhancement performance, but also is very efficient in processing speed; the dark part enhancement performance is excellent, and meets the requirements of dark part target recognition and dark part information collection and acquisition in the airborne scene. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the application, the drawings needed in the embodiment description will be briefly introduced as follows.

[0022] Figure 1 It is a flow chart of the fast airborne dark light image enhancement method in the specific embodiment;

[0023] Figure 3 It is a curve graph of the gamma adaptive change and the traditional gamma transformation in the specific embodiment;

[0024] Figure 2 It is a comparison graph of the adaptive gamma transformation and the traditional gamma transformation in the specific embodiment; ) in the specific embodiment;

[0025] Figure 4 It is a comparison result graph of the input image enhancement by the existing benchmark method and the method of the application in the specific embodiment. DETAILED DESCRIPTION

[0026] The advantages and characteristics of the application will be more clearly understood with the following description of specific embodiments. However, these embodiments are only exemplary and do not constitute any limitation on the scope of the application. Those skilled in the art should understand that the details and forms of the technical solutions of the application can be modified or replaced without departing from the spirit and scope of the application, and such modifications and replacements all fall within the protection scope of the application.

[0027] In order to overcome the shortcomings that the calculation efficiency and performance cannot be compatible when the brightness of the dark light image is enhanced by using the traditional method, the embodiment discloses a fast airborne dark light image enhancement method. The fast airborne dark light image enhancement method is described below through specific examples, and reference is made to Figure 1 The fast airborne dark light image enhancement method includes the following steps:

[0028] Step 1: The input RGB image is converted to the HSV space through color space transformation, and the V channel brightness component is extracted in the HSV space (that is, a single channel image ), and the method proposed in the present application is implemented on the single channel image , that is, the brightness enhancement is implemented on the dark light area with insufficient brightness.

[0029] Step 2: The edge-preserving smoothing processing is implemented on the single channel image , and the edge-preserving smoothing image is obtained , wherein represents the edge-preserving smoothing operator. Since there are many edge-preserving smoothing operators, such as the bilateral filter (Bilateral Filter), the guided image filter (Guided Image Filter), the weighted guided image filter (Weighted Guided Image Filter) and the like, the guided image filter is adopted as the edge-preserving smoothing operator in the present application in consideration of the calculation efficiency and the edge-preserving performance.

[0030] Step 3: The light response intensity of the sensor at the object boundary in the dark light environment is weaker than that in the normal light,

[0031] Therefore, the high frequency information pre-enhancement processing is implemented on the dark light image , and the pre-enhanced image is obtained , wherein is the high frequency information pre-enhancement coefficient.

[0032] Step 4: The contrast-enhanced dark light image is obtained through the high-order fractional model, and the high-order fractional model is as follows: , wherein is a positive integer greater than 1.

[0033] Although the gamma transformation can improve the brightness directly, the contrast is reduced. The contrast of the dark light image can be improved by using the high-order fractional model, so that the contrast lost by the image after the gamma transformation is offset, and thus the excellent visual effect is obtained. Reference is made to Figure 2As shown in the figure, (a) shows the original dark image, (b) shows the traditional gamma transformation, and (c) shows the result of the present invention after steps 2 and 4, and the traditional gamma transformation. Furthermore, the present invention proposes a higher-order fractional model ( Excellent performance (under certain circumstances);

[0034] Step 5: Since directly applying the traditional gamma transform method results in overexposure of bright areas and insufficient enhancement of dark areas, this invention proposes an adaptive gamma transform method to mitigate the shortcomings of the traditional adaptive gamma transform.

[0035] Specifically, this invention utilizes the adaptive gamma transform to enhance the contrast of low-light images. Brightness enhancement yields V channel components. Dark-light enhancement image for: ,in, These are the adaptive gamma transform coefficients proposed in this invention.

[0036] In this invention, let ,when When the pixel value is close to 0, g approaches 0.3, achieving a better effect on improving shadow brightness than traditional gamma transform; when When the pixel value is close to 1, A value close to 1.3 can achieve better brightness suppression in bright areas than traditional gamma transform, thus preventing overexposure. See also Figure 3 The figure shows the curves of the traditional gamma transform and the adaptive gamma transform proposed in this invention. As can be seen from the figure, the adaptive gamma proposed in this invention significantly enhances dark areas. In airborne scenarios, better enhancement of dark areas is beneficial for detecting hidden targets.

[0037] For airborne scenarios, the runtime and shadow enhancement effect of the method are equally important. The effectiveness and efficiency of the proposed method can be demonstrated by some high-performance benchmark methods. Common comparison methods include: histogram-based layer difference representation (LDR) method, robust Retinex-based RRM method, Retinex-based structure and texture-aware STAR method, and Zero-DCE++ method based on no-reference depth curve estimation.

[0038] This embodiment selected two sets of RGB color low-light images with a pixel size of 1280x720 pixels. Through testing on the MATLAB platform, the running time of the method of this invention and the traditional method for processing the above low-light images was compared. The results are shown in the table below, which presents the comparison of the running time between the method of this invention and the traditional model-based method:

[0039] Table 1

[0040]

[0041] From the above table, it can be seen that the method adopted by the present application has a much lower time than the time of the traditional method when processing dark light images, and has higher efficiency, and the experimental comparison results of the above reference method and the method of the present application are shown in Figure 4

[0042] The fast airborne dark light image enhancement method disclosed in the specific embodiment obtains a single-pass luminance image through color space conversion; performs edge-preserving smoothing on the luminance image; pre-enhances high-frequency information of the luminance image through the edge-preserving smoothed luminance image; and finally, obtains the final airborne dark light enhanced image through the high-order fractional model and the adaptive gamma transformation proposed in the present application. In the airborne dark light scene, compared with the existing reference method, the present application not only has high execution efficiency, but also achieves very excellent dark light enhancement performance.

[0043] The above only describes the preferred embodiments of the present application and is not intended to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0044] In addition, it should be understood that although the present specification is described in terms of embodiments, not every embodiment contains only one independent technical solution, and the specification is described in this way only for the sake of clarity, and those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be properly combined to form other embodiments that those skilled in the art can understand.​

Claims

1. A fast on-board dim-light image enhancement method, characterized in that, The method comprises the following steps: inputting a dark-light image, performing color conversion to an HSV color space; extracting a V channel component of the dark-light image after color conversion to obtain a single-channel image; Filter the single-channel image based on the edge-preserving smoothing operator to obtain an edge-preserving smoothing image ; performing high frequency information pre-enhancement on the edge-preserving smoothing image to obtain a pre-enhanced image performing high frequency information pre-enhancement on the edge-preserving smoothing image to obtain a pre-enhanced image ; based on a high-order fractional model, enhancing contrast of the pre-enhanced image to obtain a contrast-enhanced dark-light image wherein the high-order fractional model is , a positive integer greater than 1.​ based on an adaptive gamma transformation method, enhancing the brightness of the dark-light image to obtain a dark-light enhanced image.

2. The rapid on-board dim light image intensification method of claim 1, wherein: the pre-enhanced image is wherein, is a high frequency information pre-enhancement coefficient, I is a single channel image.

3. The rapid on-board dim light image intensification method of claim 1, wherein: The dark-light enhanced image is g is an adaptive gamma coefficient, 0 ≤ g ≤ 1.

Images