Issn 2091-5446 ilmiy axborotnoma научный вестник scientific journal
Download 1.19 Mb. Pdf ko'rish
|
ilovepdf merged
- Bu sahifa navigatsiya:
- ILMIY AXBOROTNOMA INFORMATIKA 2021 - yil, 1 - son
Descriptor Computation: In this stage, the goal is to create descriptive for the patch that is compact,
highly distinctive and to be robust to changes in illumination and camera viewpoint. The image gradient magnitudes and orientations are sampled around the keypoint location. These values are illustrated with small arrows at each sample location as in Fig (1). In order to achieve orientation invariance, the coordinates ILMIY AXBOROTNOMA INFORMATIKA 2021 - yil, 1 - son 109 of the descriptor and the gradient orientations are rotated relative to the keypoint orientation. In our implementation, a 16 × 16 sample array is computed and a histogram with 8 bins is used The Number of features depends on image content, size, and choice of various parameters such as patch size, number of angels and bins, and peak threshold. These parameters will be briefly described below. Peak Threshold (PeakThr) parameter is used to determine the dimension of feature vectors because PeakThr represents the amount of contrast to extract a keypoint. The optimum value of PeakThr is 0.0 because when the value of PeakThr parameter increased, the number of features decreased and more keypoints are eliminated; thus, the robustness of feature matching will decreased [4]. The patch size (Psize) parameter is used to extract different fine grained of features. Increasing the size of patches will decrease the dimension of feature vector, but at the same time it increases CPU time. Fig.1. Keypoints and matching between two different characters of Uzbek sign language images, (a) Keypoints or Extrema of SIFT, (b) Matching between two images based on SIFT features The feature vector at each keypoint is bins×bins×angels. The problem is to experiment many patch sizes to reach to the Psize that gives better accuracy. As reported in [4], increasing Psize of SIFT will produce global features. Also, decreasing Psize will not extract enough features for identification. Using different number of angels (Nangels) and number of bins (Nbins) will collect features in different orientations. Increasing the Nangels increases the number of features hence improve the accuracy of the system. on the other hand decreasing Nangels and Nbins leads to small number of features and the identification rate will decreased specially when the images are rotated because the features became variant against any rotation[11]. Download 1.19 Mb. Do'stlaringiz bilan baham: |
Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2024
ma'muriyatiga murojaat qiling
ma'muriyatiga murojaat qiling