Unmanned aerial vehicle image acquisition enhancement method

Abstract

The unmanned aerial vehicle image acquisition enhancement method can effectively improve the local brightness of the image, improve the image contrast and retain the contour edge texture information. The method comprises the following steps: using a Gaussian blur function to remove the high-frequency part of the image to achieve the purpose of image smoothing for the collected low-illumination image; converting the low-illumination image from an RGB color space to a YUV color space and extracting a Y brightness component; dividing bright and dark areas according to the calculated brightness coefficient; calculating a correction coefficient according to the pixel mean value of a sliding window and a set threshold value to complete the correction of the image brightness; and finally using a cubic fast mean algorithm to replace Gaussian filtering for denoising to reduce the time complexity. The image detail definition is improved, and the color fidelity effect is achieved.

Description

Technical Field

[0001] This invention relates to the field of computer image processing technology, and more particularly to a method for enhancing images acquired by a drone. Background Technology

[0002] With the continuous development of digital image processing and computer vision, the application of drones for image acquisition via cruise is becoming increasingly common. Low-light images acquired by drones under uneven lighting conditions exhibit low illumination characteristics, significantly hindering efficient post-processing and greatly inhibiting the application and development of drones in the field of image detection. While existing image enhancement algorithms have shown significant improvement in image quality under foggy and nighttime shooting conditions, they still haven't effectively eliminated problems such as localized underexposure, halos, and unclear image contours and textures, resulting in low processing efficiency. Summary of the Invention

[0003] In view of this, the image enhancement method for drones provided by the present invention solves the technical problem of low efficiency in low light image processing of existing methods. While maintaining and improving image quality, the present invention effectively eliminates problems such as local underexposure, halo and unclear outline texture of the image, and improves detail clarity and achieves color fidelity.

[0004] A method for enhancing images acquired by a drone, comprising:

[0005] S101: The low-light image is denoised using a Gaussian blur function to obtain the first image;

[0006] S102: The first image is converted from the RGB color space to a grayscale image, and the luminance component Y is extracted in the YUV color space of the grayscale image to obtain the histogram of the luminance component Y.

[0007] S103: Determine the bright and dark areas, wherein a preset mean pixel value is used, and pixels in the histogram whose pixel value is less than the preset mean pixel value are used as pixels in the dark area, and pixels whose pixel value is greater than or equal to the preset mean pixel value are used as pixels in the bright area.

[0008] S104: Determine the correction coefficient α, which is determined based on the pixels in the bright area;

[0009] S105: The first image undergoes a nonlinear transformation based on the correction coefficient α to obtain the second image;

[0010] S106: The second image is subjected to noise reduction processing.

[0011] Beneficial effects

[0012] This invention first uses a Gaussian blur function to denoise the acquired low-light image to smooth the image edges. Then, it divides the image into bright and dark regions, converts the RGB image acquired by the drone into a grayscale image, calculates the mean and variance of the grayscale pixels, and divides the bright and dark regions by calculating a brightness coefficient. Finally, it calculates a correction coefficient and performs brightness correction. The correction coefficient is calculated based on the pixel mean of the sliding window and a set mean point to complete the image brightness correction, and three fast mean filters are used for denoising. The results of the invented image enhancement algorithm are compared with existing SSR, MSR, and MSRCR algorithms from both subjective visual and objective evaluation metrics. The results show that the algorithm of this invention can effectively improve local image brightness, enhance image contrast, and preserve contour edge texture information. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0015] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0016] The following specific examples illustrate the implementation of this disclosure. Those skilled in the art can easily understand other advantages and effects of this disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. This disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0017] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using other structures and / or functionalities besides one or more of the aspects set forth herein.

[0018] It should also be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this disclosure. The drawings only show the components related to this disclosure and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0019] Furthermore, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that these aspects can be practiced without these specific details.

[0020] See Figure 1 The present invention provides an image enhancement method for drones, adapted for processing low-light images. The method includes:

[0021] S101: The low-light image is denoised using a Gaussian blur function to obtain the first image. Specifically...

[0022] Low-light images are processed using Gaussian blur convolution kernels in a multi-kernel parallel manner. Specifically, the Gaussian blur convolution kernel is applied to the low-light image, performing sliding operations from top to bottom and from left to right. The expression is as follows:

[0023]

[0024] Where (x, y) represents the pixel coordinates of the target image; G(x, y) represents the pixel value after the convolution operation; r represents the radius of the convolution kernel; s(x, y) represents the initial pixel value of the low-light image; and f(j, k) represents the pixel value of the convolution kernel at coordinates (u, v), thus obtaining the first image.

[0025] S102: The first image is converted from the RGB color space to a grayscale image. The luminance component Y is extracted from the YUV color space of the grayscale image to obtain its histogram. Specifically...

[0026] The preset average pixel value is 128;

[0027] The expression for extracting the luminance component Y from the first image after converting it to the YUV color space is as follows:

[0028] When performing image transformation on the component map of the luminance component Y, a mapping function is used to transform the Y component map matrix into the target image matrix after luminance adjustment.

[0029] p after (i,j)=f(p before (i,j))

[0030] Where, p after (i,j) represents the pixel value after brightness adjustment; p before (i,j) represents the pixel value before brightness adjustment; f(p before (i,j)) represents the transformation of pixel values ​​before brightness adjustment using a mapping function, resulting in a histogram.

[0031] S103: Determine the bright and dark areas, wherein the preset mean pixel value is used. Pixels with pixel values ​​less than the preset mean pixel value in the histogram are classified as dark area pixels, and pixels with pixel values ​​greater than or equal to the preset mean pixel value are classified as bright area pixels.

[0032] S104: Determine the correction factor α. The correction factor α is determined based on the pixels in the bright area. Specifically...

[0033] The pixels in the bright area are represented as A(x) i ,y i The pixel average value is calculated by comparing the pixel value with the preset mean pixel value to obtain the correction coefficient α required for the transformation. The expression is as follows:

[0034]

[0035] Where (u, v) represents the pixel coordinates of the mean point 128; n represents the number of pixel values ​​in the bright or dark region; x i and y i Let A be the coordinates of pixel A.

[0036] S105: The first image undergoes a nonlinear transformation based on the correction coefficient α to obtain the second image. Specifically, the first image undergoes a nonlinear transformation, expressed as O = R. a ×C, where,

[0037] R represents the pixel value before brightness adjustment; C represents the brightness scaling factor; O represents the pixel value after non-linear transformation; and a is the correction factor.

[0038] S106: The second image undergoes denoising processing, specifically...

[0039] The second image is denoised using a fast mean filter, where...

[0040] Fast mean filter Sum[i+1] = Sum[i] + x[i+r+1] - x[i-r], where Sum represents the summation operation, r represents the filter radius, i represents the i-th window, and x represents the element being filtered.

[0041] The summation operation is performed on the (i+1)th window. Sum[i+1] can be calculated by combining the sum of the ith window, Sum[i], with the leftmost x[i-r] and the rightmost x[i+r+1] to obtain the third image. This method of obtaining the image is independent of the size of the filter.

[0042] S107: The third image is fed into a neural network model for training, such as a v5s model. The method of this invention enables effective identification of targets from low-light images acquired by the UAV under uneven lighting conditions.

[0043] Compared to traditional image enhancement methods, this invention employs a nonlinear transformation approach to preserve image contrast and maintain color accuracy. It also utilizes a sliding window to accelerate computation, resulting in faster processing speeds. This invention implements an image enhancement method for drone-captured images, significantly improving image quality under foggy and nighttime shooting conditions. Furthermore, the enhanced images obtained using this method effectively increase information density.

[0044] The above are merely specific embodiments of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this disclosure should be included within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.

Claims

1. A method for enhancing images acquired by a UAV, adapted for processing low-light images, characterized in that, Its methods include, S101: The low-light image is denoised using a Gaussian blur function to obtain the first image. The low-light image is processed using a Gaussian blur convolution kernel in a multi-kernel parallel manner. The Gaussian blur convolution kernel is applied to the low-light image, performing sliding operations from top to bottom and from left to right. The expression is as follows: , where (x,y) represents the coordinates of the pixel; G(x,y) represents the pixel value in the first image after the convolution operation; r represents the radius of the convolution kernel; s(x,y) represents the initial pixel value of the low-light image; f(j,k) represents the pixel value of the convolution kernel at coordinates (u,v); S102: The first image is converted from the RGB color space to a grayscale image, and the luminance component Y is extracted in the YUV color space of the grayscale image to obtain the histogram of the luminance component Y. S103: Determine the bright and dark areas, wherein a preset mean pixel value is used, and pixels in the histogram whose pixel value is less than the preset mean pixel value are used as pixels in the dark area, and pixels whose pixel value is greater than or equal to the preset mean pixel value are used as pixels in the bright area. S104: Determine the correction coefficient α, which is determined based on the pixels in the bright area; S105: The first image undergoes a nonlinear transformation based on the correction coefficient α to obtain a second image, wherein... The first image undergoes a nonlinear transformation, expressed as follows: ,in, R represents the pixel value before brightness adjustment; C represents the brightness scaling factor; O represents the pixel value after non-linear transformation; α is the correction factor. S106: The second image undergoes denoising processing, wherein... The second image is denoised using a fast mean filter, where... Fast mean filter Sum[i+1]=Sum[i]+x[i+r+1]-x[ir], where Sum represents the summation operation, r represents the filter radius, i represents the i-th window, and x represents the element being filtered; The summation operation is performed on the (i+1)th window. Sum[i+1] can be calculated by combining the sum of the ith window, Sum[i], with the elements of the leftmost x[ir] and the rightmost x[i+r+1], thus obtaining the third image.

2. The image enhancement method for UAV acquisition according to claim 1, characterized in that, Step S102 includes, The preset mean pixel value is 128; The expression for extracting the luminance component Y from the first image after converting it to the YUV color space is as follows: When performing image transformation on the component map of the luminance component Y, a mapping function is used to transform the Y component map matrix into the target image matrix after luminance adjustment. in, This represents the pixel value after brightness adjustment; This represents the pixel value before brightness adjustment; This means that the pixel values ​​before brightness adjustment are transformed using a mapping function, and the resulting histogram of the brightness component Y is obtained.

3. The image enhancement method for UAV acquisition according to claim 2, characterized in that, Step S104 includes, The pixels in the bright area are represented as The pixel average value is calculated by comparing it with the preset mean pixel value to obtain the correction coefficient α required for the transformation, expressed as follows: ， In the formula, (j,k) represents the pixel coordinates of the mean point 128; n represents the number of pixel values ​​in the bright or dark area; and xi and yi are the coordinates of pixel A.

Images