| Research Abstract |
New nonlinear digital signal processing methodologies, which are supposed to be highly demanded in our multimedia world, are proposed. For instance, in multimedia communication environment, we often transmit and show our face images on the displays of communication terminals. Here face images contain some undesirable rugged components such as wrinkles and spots on skin, which people wish to hide away, as well as noises and block artifacts entering from the communication network. In this research, a new nonlinear digital filter is proposed, which can remove such undesirable components quite effectively. Especially, this filer is powerful in showing the skin to be beautiful. In conventional ways, linear low-pass filters are used to remove such noisy components, but if we apply it, sharp edges are also removed as well as the noisy components and the image becomes blurred. Thus, a nonlinear method is adopted in this system. Moreover, in order to apply for real time processing of moving images on mobile terminals, the algorithm is kept simple and fast in realization.
This system is shown to be effective for actual images in reducing noises and block artifacts, and also the undesirable ruggedness of skins. Moreover, in order to implement this system efficiently, a new idea of NDSP (nonlinear digital signal processor) is proposed, which is realized by adding simple nonlinear functions to usual DSP.
This research can be also applied to the problem of sharpening blurred images, often taken with digital cameras. A new NDSP system, realized as a network of such NDSP's, is proposed. This system can sharpen the blurred images without amplifying small noisy components on images.
|
