下載stirmark benchmark 4.0:我是下載binary and C/C++ source code並解壓縮。
目錄說明:
/ | Bin/ | Benchmark/ | the benchmarking tool itself | ||
Libraries/ | libraries sent by users | ||||
Profiles/ | profiles are text configuration files describing the test to be applied (one profile per application and per 'robustness' level) | ||||
Media/ | Input/ | Images/ | Set1/ | 輸入的影像放這裡 | |
Set2/
… MyFolder/ |
... | ||||
Output/ | (Same substructure is created for input. ) | 輸出影像的資料夾 | |||
Step 1.將嵌入浮水印的影像(watermarked image)放:Media/ Input/ Images/ Set1/
Step 2.到Profiles/下找SMBsettings.ini:改攻擊參數。可不改,用預設。
Step 3.到Bin\Benchmark\下找StirMark Benchmark.exe,點二下。此為執行檔。
Final Step:執行後,所有攻擊後的圖在:Media/ Output/ Images/ Set1/中。
stirmark benchmark 4.0 | 破壞程度(小) | ---> | 破壞程度(大) | example(預設) | 說明 | 問題 | 使用狀況 |
Test1=Test_PSNR | Strength of the watermark embedding | ||||||
Test2=Test_EmbedTime | ; Number of embeddings with random key per media ; This is used to compute the average embeding and extraction time | ||||||
Test3=Test_AddNoise | 0 | -> | 100 | list=5 | |||
Test4=Test_JPEG | 100 | -> | 0 | list=5 | JPEG_5:等級為5的JPEG壓縮。(等級低,壓縮比高)。 | OK | |
Test5=Test_MedianCut | 3 | 5、... | the odd number < min(width px, height px) | list=3 | 中位數濾波攻擊。MEDIAN_3:採用3X3的遮罩。 | OK | |
Test6=Test_ConvFilter1(Gaussian filtering) Convolution |
filter1 = 3 3 9 1 2 1 2 4 2 1 2 1 | 和高斯函數做convolution。也就是該點像素以高斯函數做為附近像素值的加權平均之權重,加權後的均值取代原該點像素。所以有平滑、模糊的效果。 3的遮罩為[1 2 1;2 4 2;1 2 1]。 參考:我+單元五、空間濾波 |
平滑...為什麼我的會變白~~~*2 | OK | |||
Test6=Test_ConvFilter2(sharpening) | filter2 = 3 3 9 0 1 0 1 5 1 0 1 0 | 3的遮罩為[0 1 0;1 5 1;0 1 0]。 參考:WEB化的Watermark安全性 |
銳化...為什麼我的會變黑~~~*2 | OK | |||
Test7=Test_SelfSimilarities | 使用RGB色系,RGB通道,交換攻擊 | ||||||
Test8=Test_RemoveLines | |||||||
Test9=Test_Cropping | 100 | -> | 0 | list=50 | 取中間部分。50:表示取下中間長寬50%做為測試 | OK | |
Test10=Test_Rescale | 100 | -> | 0 or infinite | list=10 | 縮放百分比:10% | OK | |
Test11=Test_Rotation | 0 | -> | 360 | list=0.1 | 逆時針轉0.1度 | OK | |
Test12=Test_RotationCrop | |||||||
Test13=Test_RotationScale | |||||||
Test14=Test_Affine | Affine | ||||||
Test15=Test_SmallRandomDistortions | |||||||
Test16=Test_LatestSmallRandomDistortions | |||||||
PS:
*1:StirMark出來的圖檔是BMP格式。但進去格式不分。
*2:Test_ConvFilter用法
[Test_ConvFilter] ; filter coefficients
; first two numbers correspond to width and height
; second number is division factor
; Gaussian filtering
filter1 = 3 3 16 1 2 1 2 4 2 1 2 1
;filter1 =mask的寬 mask的高 除數 1 2 1 2 4 2 1 2 1
;這裡除數取1+ 2 +1+ 2+ 4+ 2+ 1+ 2+ 1=16
; sharpening
filter2 = 3 3 1 0 -1 0 -1 5 -1 0 -1 0
; edge
filter3 = 3 3 1 0 -1 0 -1 4 -1 0 -1 0
沒有留言:
張貼留言