Underwater image enhancement method based on deep learning
An underwater image and deep learning technology, applied in the fields of deep learning and computer vision, can solve the problems of low illumination, inconspicuous contrast, and low resolution of underwater images, so as to solve serious image fogging and improve generalization. and robustness, the effect of visual quality improvement
Pending Publication Date: 2022-05-24
UNIV OF ELECTRONICS SCI & TECH OF CHINA
View PDF11 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0005] The present invention provides an underwater image enhancement method based on deep learning, the purpose of which is to improve the overall image quality of underwater images obtained in the prior art, such as low illumination, high noise, low resolution, serious fogging, inconspicuous contrast, and color distortion. Visual quality degradation issues, avoid further affecting the recognition, tracking and detection of target objects
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0054] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0055] The overall flow of the method of the present invention is as follows figure 1 shown, which includes the training process of the neural network as follows figure 2 As shown, the multi-scale degradation feature extraction process is as follows image 3 shown. The specific construction steps of the underwater image enhancement in the embodiment of the present invention are as follows:
[0056] Step 1, channel splitting: split the input original image (such as Figure 4 shown) is divided into three channel matrices R, G, B, such as Figure 5 , Image 6 and Figure 7 shown, and normalize the pixel values of the three channel matrices to between 0-1;
[0057] Step 2, multi-scale degradation feature extraction: input the three channels obtained in step 1 into the three multi-scale fusion attention modules to extract the underwater decay feature...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention discloses an underwater image enhancement method based on deep learning, and belongs to the field of deep learning and computer vision. According to the method, an input underwater image is divided into R, G and B color channels, attenuation feature maps of the different color channels are learned by using a multi-scale fusion attention module, channel stacking is performed on the attenuation feature maps of the different color channels, interested attenuation features are extracted by using a channel attention mechanism, and the multi-scale fusion attention module is used for learning the attenuation feature maps of the different color channels. And finally, reducing the output dimension of the feature map by using 1 * 1 convolution, fusing the feature map with the input image, and enhancing the input underwater image. The method provided by the invention can effectively improve the overall visual quality of the underwater image, and has good robustness for the underwater images of various scenes shot under the condition of attenuated illumination. The problem that the visual quality of an image obtained by a traditional underwater image enhancement method is not obviously improved due to large color correction chromatic aberration, low image illumination, high noise, serious image atomization and the like is solved.
Description
technical field [0001] The invention belongs to the fields of deep learning and computer vision, in particular to an underwater image enhancement method based on computer vision. Background technique [0002] High-quality underwater images are the premise of understanding underwater information, and have a very important impact on actual seabed exploration, underwater biodiversity research, underwater environmental protection and other engineering projects and scientific research. However, affected by the physical environment such as impurities and air bubbles in the water, the quality of underwater images often degrades to a certain extent. The scattering effect of water on light causes the acquired underwater image texture to be blurred, and there is a certain fogging phenomenon. The absorption of light by water causes the color distortion of the underwater image. The longer the wavelength of light is, the more obvious the absorption is, resulting in the underwater image ...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06V10/25G06K9/62G06N3/04G06N3/08G06T5/00G06T7/90G06V10/82G06V10/80
CPCG06N3/084G06T7/90G06T2207/10024G06N3/045G06F18/253G06T5/77G06T5/70Y02A90/30
Inventor 左琳徐冯杰张昌华刘宇杨兴刘斌胡建罗茂林
Owner UNIV OF ELECTRONICS SCI & TECH OF CHINA