Processing of face images is an important branch of machine vision. In real scenarios, the quality of the acquired images often does not reach the ideal condition therefore producing wrong results. While face super-resolution and rotation are two different ways to enhance the quality of face images, these two methods have always existed independently. The super resolution of face is to improve the image quality and get high resolution image from low resolution face image, and the rotation of face is to get the image under different view of the target and enrich the target person information. Both techniques have different approaches, but both are designed to improve the image quality for easier processing such as classification, detection and recognition. Therefore, can we combine the two methods, convert an image from a low-resolution image to a high-resolution image, and then use face rotation to obtain images of other views of the target face. We can obtain a higher quality image compared with the two independent methods. In this paper, we use a supervised learning method for face super-resolution and a self-supervised face rotation method for combining experiments, and the results show the reliability of combining the two methods.
KEYWORDS: Holograms, Denoising, Education and training, Speckle, Image processing, Holography, Digital holography, Histograms, Deep learning, 3D image reconstruction
Digital holographic microscopy (DHM) is a non-contact and high accuracy measurement technique widely used in biomedicine, microstructure,and other fields.The quality of the reconstructed image and the effectiveness of holographic microscopy were easily affected by speckle noise. Inspired by the idea of Noise2Noise, we propose a self-supervised noise2noise hologram speckle noise removal method. From the holograms that need denoising to generate the input and labels with the same noise distribution to form a training pair for training. Solve the problem that clean holograms are difficult to obtain.The training sets of this self-supervised method are generated from the holograms to be processed. As such, it avoids the need of collectting a large number of training sets. The proposed method is therefore less vulnerable to the background noises and more convenient and reliable for practical hologram speckle denoising applications.
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.