An end-to-end dehazing network with transitional convolution layer
- PDF / 2,456,516 Bytes
- 21 Pages / 439.37 x 666.142 pts Page_size
- 77 Downloads / 184 Views
An end-to-end dehazing network with transitional convolution layer Shuying Huang1,3 · Hongxia Li1 · Yong Yang2 · Bin Wang1 · Nini Rao3 Received: 23 September 2019 / Revised: 9 March 2020 / Accepted: 19 March 2020 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Image dehazing is a challenging task of restoring a clear image from a haze-polluted image. However, most popular dehazing methods have color shift and overexposure problems owing to the inaccurate estimation of the transmission and atmospheric light in the atmospheric scattering model. In view of the existing problems, this paper proposes a simple but effective dehazing network by learning the residual image between the hazy image and haze-free image. A novel transitional convolution structure is constructed in this network, which contributes to utilizing shallow structure information to enhance the final residual image. Furthermore, to provide images with more realistic color information after dehazing, a novel color difference loss based on CIEDE2000 is designed as one term of the total loss function. In addition, for some real-world images that affect practical applications, a brightness enhancement module is also introduced to restore the luminance of the images. Experiments on synthetic datasets and real-world images demonstrate that the proposed method has clear advantages in both subjective and objective evaluation indicators compared to several existing advanced algorithms. Keywords Image dehazing network · Color difference loss · Brightness enhancement module · Transitional convolution layer
1 Introduction Usually due to the limited visibility of the atmosphere in the haze weather, the images captured have low contrast and dark brightness. It is mainly because atmospheric light is scattered and reflected by aerosol particles suspended in the air, leading to attenuating the irradiance
B
Yong Yang [email protected]
1
School of Software and Internet of Things Engineering, Jiangxi University of Finance and Economics, Nanchang 330032, China
2
School of Information Technology, Jiangxi University of Finance and Economics, Nanchang 330032, China
3
School of Life Science and Technology, University of Electronic Science and Technology of China, Chengdu 610054, China
123
Multidimensional Systems and Signal Processing
Fig. 1 Examples of single image haze removal. a The hazy images. b Dehazed results of the proposed method
received by the camera lens (Wang and Yuan 2017; Sharma and Chopra 2014). Figure 1 shows some examples of haze removal for hazy images. Figure 1a shows the original hazy images, and Fig. 1b shows the corresponding dehazing results using the proposed method. As can be seen from Fig. 1a, a layer of haze veil exists on the surface of a hazy image, which makes the image blurred and some detail information hidden to some extent. However, some high-level visual tasks, such as outdoor monitoring and target recognition, have very high requirements for image quality. Therefore, it is necessary to recover th
Data Loading...
