Algorithm: Noise images filtration algorithm | |
---|---|
Input: A dataset \(D=\{{x}_{1},{x}_{2},\dots ,{x}_{i},\dots ,{x}_{n}\}\). K: the optimal K value after K-means clustering, \(n\): the number of data, \(N\): the number of CICH noise images Output: CICH noise image point set | |
1: | Extracted representation features \(\,E=\{{e}_{1},{e}_{2},\dots ,{e}_{i},\dots ,{e}_{n}\}= Extractor\left(D\right)\) |
2: | Centroid point set \(\,C=\{{c}_{1},{c}_{2},\dots ,{c}_{j},\dots ,{c}_{K}\}=Kmeans(E,K)\) |
3: | For each cluster \(\,{Q}_{j}\) in \(\,Q=\{{Q}_{1},{Q}_{2},\dots ,{Q}_{j},\dots ,{Q}_{K}\}\) do |
4: | \(\,\,{Radius}_{{Q}_{j}}=radius ({Q}_{j})\) |
5: | End for |
6: | If the number of points in \(\,{Q}_{j}>N\): |
7: | Â Â For each point \({e}_{i}{\in Q}_{j}\) do |
8: | Â Â Â If \(distance\,({e}_{i},{c}_{j})\) > \({Radius}_{{Q}_{j}}\) then |
9: | Â Â Â Â Add \(\,{e}_{i}\) to \(\,S\) |
10: | Â Â Else |
11: | Â Â Â Â \(Continue\) |
12: | Â Â End for |
13: | Else |
14: | Â Â For each point \(\,{e}_{i}{\in Q}_{j}\) do |
15: | Â Â Â Add \(\,{e}_{i}\) to \(\,S\) |
16: | Â Â End for |
17: | For each point \(\,{e}_{i}\,\) in \(\,S\,\) do |
18: | Â Â Calculate \(\,CBLOF\,({e}_{i})\) |
19: | End for |
20: | The top-\(N\) points are the noise images based on the \(CBLOF \left({e}_{i}\right)\) value ranking results |