Haze significantly impacts various fields, such as autonomous driving, smart cities, and security monitoring. Deep learning has been proven effective in removing haze from images. However, obtaining pixel-aligned hazy and clear paired images in the real world can be challenging. Therefore, synthesized hazed images are often used for training deep networks. These images are typically generated based on parameters such as depth information and atmospheric scattering coefficient. However, this approach may cause the loss of important haze details, leading to color distortion or incomplete dehazed images. To address this problem, this paper proposes a method for synthesizing hazed images using a cycle generative adversarial network (CycleGAN). The CycleGAN is trained with unpaired hazy and clear images to learn the features of the hazy images. Then, the real haze features are added to clear images using the trained CycleGAN, resulting in well-pixel-aligned synthesized hazy and clear paired images that can be used for dehaze training. The results demonstrate that the dataset synthesized using this method efficiently solves the problem associated with traditional synthesized datasets. Furthermore, the dehazed images are restored using a super-resolution algorithm, enabling the obtainment of high-resolution clear images. This method has broadened the applications of deep learning in haze removal, particularly highlighting its potential in the fields of autonomous driving and smart cities.
The complex amplitudes are usually extracted by windowing the Fourier spectra in the off-axis digital holography. However, it often needs manual operation. In order to obtain the complex amplitude automatically from a single off-axis hologram, the spatial carrier frequency phase shifting method is used in this work. Based on the fact that the complex amplitude to retrieve usually vary slowly and the carrier frequency keeps constant, three consecutive phase-shifting interferograms are obtained using the neighboring pixels along the same line in the hologram. Then the complex amplitude is calculated out. Without Fourier transforms, the computational complexity here is much smaller than that of the traditional one. Simulative and experimental results show that this method can really improve the processing efficiency during the complex amplitude reconstruction, and it can be used for automatic application of the digital offaxis holography.
Windowing the Fourier spectra is usually used to filter the zero-order and twin image in off-axis digital holography (DH). However, during the filtering process, manual operation is often needed. In our work, we present an automatic filtering method. The filtered zone is a circle in the Fourier domain. The point with the highest intensity in the +1st order is chosen as its center, the location of which is related to the angle between the reference and the objective beams. Considering that the size of the zero-order spectra is twice of that of the +1st order, the radius of the circle is set to be a third of the distance between the circle center and the very center of the Fourier spectra. With this filtering method, the desired spectra can be filtered out automatically and noniteratively. This method has been tested and validated on several samples. Compared with the iterative filtering method, it spends about 2‰ of the processing time and obtains the reconstructed images with high similarity. This approach can result in real-time use of the off-axis DH and holographic microscopy.
For digital holographic applications, we have studied the refocus criteria operating on the high-pass filtered and the nonfiltered reconstructed complex amplitudes (RCAs), respectively. A set of 10 criteria have been investigated for applications on phase, amplitude, and mixed objects in off-axis digital holography. Without filtering, the criteria always exhibit opposite behaviors for amplitude and phase objects and cannot work normally for mixed ones. With high-pass filtering on the RCAs, most of the criteria can be used regardless of the phase or amplitude nature of the objects without noises. But, in real applications with noises, they show different robustness. The investigations indicate that the criterion based on the integrated amplitude shows good exactitude and strong robustness, exhibiting a great potential in real applications, especially for multiobject systems.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.