This paper describes the novel evolution approach to direct spatio-temporal filtering. It is based on the evolution programming for image sequence filter designing. The main new element in the design of the chromosome is that it is three-dimensional. Each gene represents the filter weight. The proposed filters are tested by the real image sequences.
PL
Artykuł opisuje nowe ewolucyjne podejście do bezpośredniej filtracji czasowo - przestrzennej. Opiera się ono na zastosowaniu programowania ewolucyjnego do projektowania filtrów sekwencji obrazów. Nowym elementem jest zastosowanie trójwymiarowych chromosomów, których każdy gen reprezentuje wagę filtru. Proponowane filtry są testowane z wykorzystaniem rzeczywistych sekwencji obrazów cyfrowych.
2
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Old movies suffer from various types of degradation: severe noise, blurred edges of objects (low contrast), scratches, spots, etc. Finding an efficient denoising method is one of the most important and one of the oldest problems in image sequence processing. The crucial thing in image sequences is motion. If the motion is insignificant, then any motion noncompensated method of filtering can be applied. However, if the noise is significant, then this approach gives most often unsatisfactory results. In order to increase the quality of frames, motion compensated filters are usually applied. This is a very time consuming and awkward approach due to serious limitations of optical flow methods. In this paper, a review of various filters with motion detection when applied to the processing of image sequences coming from old movies is presented. These filters are nonlinear and based on the concept of multistage median filtering or mathematical morphology. Some new filters are proposed. The idea of these new filters presented here is to detect moving areas instead of performing full estimation of motion in the sequence and to apply exclusively 2D filters in those regions while applying 3D motion noncompensated filters in static areas, which usually significantly reduces the computational burden.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.