High-speed automatic edge detection using pixel group statistics and fuzzy-based automatic thresholding = 화소군의 통계적특성과 퍼지추론식 문턱치 결정을 이용한 윤곽선의 고속 자동추출

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 402
  • Download : 0
We present a new edge detection method, which automatically determines the edges in practical vision applications without any manual adjustment. We compute the edge magnitude and orientation of a pixel with minimal smoothing across the edges by determining the most probable edge pattern for the pixel at the early stage. We classify the pixels of the 3x3 window centered at the current pixel into two groups based on the intensity similarity. The shape of the classification gives the most suitable ideal binary pattern for the pixel and it can be described by an 8-bit code. The edge magnitude of the pixel is estimated by the difference between the average intensities of the two pixel groups in the window. The edge orientation is determined by referencing a lookup table, which is constructed offline a priori for all 256 ideal edge patterns. We also use the ideal binary pattern for the non-maxima suppression to suppress the pixels that have the ideal binary pattern of small edge confidence measure. Edge confidence measures of all ideal edge patterns are determined by the offline edge detection experiments using a great many training images. For automatic edge thresholding, we use a fuzzy reasoning method. We can determine multiple thresholds for an input image one for each pixel group that is constructed by using the intensity histogram. The pixels that have intensities between the significant local valleys of the histogram belong to the same pixel group. The edge threshold is determined at the level, which is higher than the most prevailing edge magnitude in the pixel group by some multiples of just-noticeable difference of the human eye. The simple statistics of edge magnitude such as mode and the mean and the pixel count of a pixel group are used as the inputs to the fuzzy reasoning. We use an 18-entry fuzzy rule-base and use MIN-MAX method and CENTROID method in the fuzzy reasoning. As we apply the fuzzy reasoning in the pixel group level and use the LUT efficie...
한국과학기술원 : 자동화및설계공학학제전공,
Issue Date
232063/325007  / 000949515

학위논문(박사)- 한국과학기술원 : 자동화및설계공학학제전공, 2003, [ vii, 150 p. ]


fast automatic edge detection; edge shape factor; 퍼지추론식 윤곽선 문턱치결정; ideal binary pixel pattern; 윤곽선고속자동검출; 윤곽선형태계수; 최적윤곽선패턴; edge thresholding by fuzzy-reasoning

Appears in Collection
Files in This Item
There are no files associated with this item.


  • mendeley


rss_1.0 rss_2.0 atom_1.0