Peihua Qiu

In our daily life, images (pictures) are everywhere. In some circumstances, the quality of images is not high enough for human interpretation (e.g., some satellite images or the images transmitted by the cable picture transmission systems in the early 1920s). Improvement of pictorial information for human perception is one of the major purposes of image processing. Since much of the information in an image is conveyed by the so-called edges (the outlines of the objects) and our eye-brain system has evolved to extract edges by preprocessing that begins right at the retina, edge detection and edge-preserving image reconstruction are two important topics in image processing, which are also the focus of my current research in this area. Images can be regarded as surfaces of the image intensity functions (representing the image brightness at each pixel). Therefore edge detection and edge-preserving image reconstruction are essentially the same problems as jump detection and jump-preserving surface fitting in nonparametric regression analysis. Part of my current research effort is to connect methodologies in these two closely related research areas. I also enjoy working on some statistical problems faced by researchers in the biomedical and health sciences. From early 1996, I was involved in a project to study animal sleep-wake patterns in response to light-dark cycles of varying lengths. Some challenging semiparametric regression models have been motivated by this project. Most recently I am working on multiple comparisons and nonparametric estimation of receiver operating characteristic (ROC) curves, raised by a consulting project to identify reliable predictors of the onset of juvenile myopia.