Deep learning has achieved great success in computer vision, natural language processing, recommendation systems and other fields. However, the models of deep neural network (DNN) are very complex, which often contain millions of parameters and tens or even hundreds of layers. Optimizing weights of DNNs is easy to fall into local optima, and hard to achieve better performance. Thus, how to choose an effective optimizer which is able to obtain network with higher precision and stronger generalization ability is of great significance. In this article, we make a review of some popular historical and state-of-the-art optimizers, and conclude them into three main streams: first order optimizers that accelerate convergence speed of stochastic gradient descent or/and adaptively adjust learning rates; second order optimizers that can make use of second-order information of loss landscape which helps escape from local optima; proxy optimizers that are able to deal with non-differentiable loss functions through combining with the proxy algorithm. We also summarize the first and second order moment used in different optimizers. Moreover, we provide an insightful comparison on some optimizers through image classification. The results show that first order optimizers like AdaMod and Ranger not only have low computational cost, but also show great convergence speed. Meanwhile, the optimizers that can introduce curvature information such as Adabelief and Apollo, have a better generalization especially when optimizing complex network.
Multi-view super-resolution refers to the process of reconstructing a high-resolution image from a set of low-resolution images captured from different viewpoints typically by different cameras. These multi-view images are usually obtained by an array of the same color cameras. However, the color cameras have color filter array to acquire color information, which reduces the quality of obtained images. To avoid color camera, and obtain higher resolution color images, we do research on a camera array which consists of interlaced different monochrome cameras and propose a new super-resolution method based on the camera array. Given that MVSR is an ill-posed problem and is typically computationally costly, we super-resolve multi-view monochrome images of the original scene via solve a regularization optimization problem consisting of a data-fitting term and three regularization terms on image, blur and cross-channel priors. The resulting optimization problems with respect to the desired image and with respect to the unknown blur are efficiently addressed by the alternating direction method of multiplier. Corresponding experimental results, conducted on a series of datasets captured by our own camera array system, demonstrate the effectiveness of the proposed method.
Binocular cameras have gained increasing attention because they can capture high-resolution images at a lower cost than monocular cameras. However, many existing binocular camera technologies typically require accurate depth estimation. To address this problem, this paper presents a new image enhancement method based on monochromecolored cameras. Our method replaces depth estimation with dense matching of feature points, thereby effectively reducing the computational complexity. After image matching, matrix completion is used to recover the color information of the monochrome image. Consequently, our method produces a high-quality image under the low-light condition. We built real image database for the experiments, and the results reveal that our method exhibits superior performance over existing methods.
The technology of binocular camera matures day by day. Compared with monocular camera, it can obtain higher resolution images at a lower cost than monocular cameras. However, existing high dynamic range methods based on images acquired by monocular camera, causing the result images to be noisy and blurry. In order to solve the problem, this paper presents a new high dynamic range method based on monochrome-color camera system. We first use the camera system to obtain multiple sets of different exposure monochrome-color image pairs, and then match the same exposure image pair. By using the color propagation methods, we combine the color information from color image with detail information from monochrome image, and obtain multiple sets of different exposures, sharper, low-noise images with more details. And finally get the result through high dynamic imaging and tone mapping. Experiments show that our method is better than the results of the classical method.
High dynamic range (HDR) images can show more details and luminance information in general display device than low dynamic image (LDR) images. We present a robust HDR imaging system which can deal with blurry LDR images, overcoming the limitations of most existing HDR methods. Experiments on real images show the effectiveness and competitiveness of the proposed method.
The state-of-the-art blind image deblurring (BID) methods are sensitive to noise, and most of them can deal with only small levels of Gaussian noise. In this paper, we use simple filters to present a robust BID framework which is able to robustify exiting BID methods to high-level Gaussian noise or/and Non-Gaussian noise. Experiments on images in presence of Gaussian noise, impulse noise (salt-and-pepper noise and random-valued noise) and mixed Gaussian-impulse noise, and a real-world blurry and noisy image show that the proposed method can faster estimate sharper kernels and better images, than that obtained by other methods.
Our paper presents a method for reconstructing a high-resolution (HR) image from a set of multi-view color images captured by a camera array. First, an accurate depth map of low-resolution (LR) image captured by a selected reference camera is obtained using graph cuts. Then, a HR image corresponding to the reference camera can be estimated by super-resolution reconstruction. Experiments on real images show the effectiveness of our method.
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.