Combined edge detection using wavelet transform and signal registration

摘要: This paper presents a novel edge detection algorithm, using Haar wavelet transform and signal registration. The proposed algorithm has two stages: (a) adaptive with the maximum entropy thresholding technique on time-scale plane (b) linkage into contour line registration in order to close discontinuities calculate confidence index for linkages. measures level of adjacent points structure. Experimenting synthetic images, we found out lower can be set approximately e^-^2. method was tested 200 images at different signal-to-noise ratios (SNRs) 11 clinical images. We assessed its reliability, accuracy robustness mean absolute distance (MAD) metric our index. results MAD yield 0.7 standard deviation (std) 0.14, while is 0.48 std 0.19 (the values are averaged over SNRs from 3 50dB each 20 Monte-Carlo runs). Our assessment where references were expert's annotations, give equal 1.36+/-0.36 (mean+/-std) 0.67+/-0.25 (mean+/-std).