Engineering Commons^™

Single Image Reflection Removal Beyond Linearity, Qiang Wen, Yinjie Tan, Jing Qin, Wenxi Liu, Guoqiang Han, Shengfeng He

Research Collection School Of Computing and Information Systems

Due to the lack of paired data, the training of image reflection removal relies heavily on synthesizing reflection images. However, existing methods model reflection as a linear combination model, which cannot fully simulate the real-world scenarios. In this paper, we inject non-linearity into reflection removal from two aspects. First, instead of synthesizing reflection with a fixed combination factor or kernel, we propose to synthesize reflection images by predicting a non-linear alpha blending mask. This enables a free combination of different blurry kernels, leading to a controllable and diverse reflection synthesis. Second, we design a cascaded network for reflection removal with …

Cloud Workload Allocation Approaches For Quality Of Service Guarantee And Cybersecurity Risk Management, Soamar Homsi

FIU Electronic Theses and Dissertations

It has become a dominant trend in industry to adopt cloud computing --thanks to its unique advantages in flexibility, scalability, elasticity and cost efficiency -- for providing online cloud services over the Internet using large-scale data centers. In the meantime, the relentless increase in demand for affordable and high-quality cloud-based services, for individuals and businesses, has led to tremendously high power consumption and operating expense and thus has posed pressing challenges on cloud service providers in finding efficient resource allocation policies.

Allowing several services or Virtual Machines (VMs) to commonly share the cloud's infrastructure enables cloud providers to optimize resource …