Research

■Convex optimization and its application to signal recovery

Although data acquisition sensors have dramatically improved for decades, we still face difficult situations when capturing clean signals. In such cases, the captured signals may suffer from noise, blur, color artifacts, and so on. Signal recovery based on optimization has recently been extensively studied to obtain desired signals from degraded ones. Our group tries to model those problems using convex optimization and design more efficient priors.

■Graph signal processing

Graph Signal Processing (GSP) is an emerging field at the intersection of signal processing, data science, and network theory. It extends traditional signal processing techniques to data indexed by graphs, which can naturally represent various complex structures in modern data science applications. The core idea is to analyze and process signals that reside on the nodes of a graph, leveraging the underlying graph structure to enhance and generalize classical signal processing methods. 

■Image and Video Coding

With the increase in image and video quality (high spatial/spectral/temporal/view resolution), the amount of information is significantly increasing. Thus, higher performance is desired for image and video codecs.  Our group is developing coding tools for both standard and non-standard image/video coders (JPEG-XR, HEVC).

■Sparse signal representation (Wavelet/Filterbank/Frames) 

In this topic, we focus on developing bases and frames to represent signals sparsely in a way that captures essential information efficiently and effectively. This research has profound implications across various fields, including data compression, image processing, medical imaging, and communications. 

■Underwater image enhancement 

Images taken underwater are usually dominated by a strong color artifact, which degrades post-processing performance, e.g., object recognition. We develop robust color correction and enhancement algorithms.