A Fast and Practical CNN Method for Artful Image Regeneration

报告人: Xiaolin Wu, Department of Electrical and Computer Engineering, McMaster University

时间: 2019年11月6日 14：30-16：30

地点：js33333金沙线路检测335会议室

摘要:

Although artists’ actions in photo retouching, or artful image regeneration, appear to be highly nonlinear in nature and very difficult to model analytically, we find that the net effects from a mundane image to its final beautified version can be characterized, in most cases, by a parametric monotonically non-decreasing tone mapping function in the luminance axis and by a linear transform in the chrominance plane.  This allows us to greatly simplify the existing CNN methods for mimicking the artists in artful image regeneration, and design a new artful image regeneration network (AIRNet). The objective of the AIRNet is to learn the image-dependent parameters of the luminance tone mapping function and the linear chrominance transform, rather than learning the end-to-end pixel level mapping as in the standard practice of current CNN methods for image restoration and enhancement.  The proposed new approach reduces the complexity of the neural network by two orders of magnitude, and as a beneficial side effect, it also improves the robustness and the generation capability at the inference stage.  The construction of the new AIRNet is made possible by an innovative technique of generating the required paired training images before and after photo retouching.

报告人简介:

Xiaolin Wu, Ph.D. in computer science, University of Calgary, Canada, 1988. Dr. Wu started his academic career in 1988, and has since been on the faculty of Western University, New York Polytechnic University (NYU Poly), and currently McMaster University. He holds the NSERC senior industrial research chair in Digital Cinema. His research interests include image processing, computer vision multimedia signal coding and communication, joint source-channel coding, multiple description coding, and network-aware visual communication. He has published over two hundred-sixty research papers and holds five patents in these fields.  Dr. Wu is an IEEE fellow, McMaster Distinguished Engineering Professor, an associated editor of IEEE Transactions on Image Processing, and served on the technical committees of many IEEE international conferences/workshops. Dr. Wu received numerous international awards and honors.