Skip to main content

Metode Otsu

Review A Threshold Selection Method from Gray-Level Histograms



Pengantar Metode Otsu

Metode otsu merupakan metode memilih batas ambang (threshold) secara otomatis dari tingkat keabu-abuan  histogram  melalui  analisis  diskriminan.  Analisis  diskriminan  tersebut diharapkan dapat  memaksimalkan  pemisahan  objek  (foreground)  dan  latar  belakang (background).
Citra memiliki  gray level  antara  rentang nilai 1 sampai L, dimana L=255. Sedangkan k merupakan nilai  ambang  yang  berada  diantara  nilai  tersebut.  Dengan  kata  lain,  background bernilai [1,,k] dan  foreground  bernilai [k+1,…,L]. Probabilitas setiap piksel, zeroth cumulative moment, first cumulative  moment, total nilai  mean, serta nilai ambang k  ditunjukkan berturutturut pada Persamaan 1, 2, 3, 4 dan 5.





Kelebihan metode otsu antara lain:

  • Metode yang diusulkan lebih sederhana karena pencarian threshold hanya memanfaatkan, zeroth cumulative moment, first cumulative moment, dan total nilai mean pada histogram
  • Menganalisa mean level kelas dan evaluasi kelas pemisahan Berdasarkan Gambar 1.e dan 1.f pada hasil penelitian Nobuyuki Otsu, terlihat jelas bahwa metode ini mampu membedakan objek dan background yang memiliki derau dengan tingkat gray level yang hampir mendekati warna objek.
  • Pemilihan optimal threshold stabil karena tidak berdasarkan nilai turunan untuk suatu kasus yang tidak global (local)




Kemahanan metode otsu antara lain:

  • Berdasarkan Gambar 3.e dan 3.f, terdapat detail objek yang hilang.
  • Objek kurang smooth


Saran perbaikan antara lain:

  • Metode  thresholding  dapat  dilakukan  dengan  mengumpulkan  cluster  pada  area  yang memiliki  similarity  sama  dan  distance  yang  saling  berdekatan  dengan  mengembangkan dendogram/tingkat  keserupaan  gray  level  dari  keabu-abuan  berdasarkan  inter  class  dan intra class sehingga pemisahan antar objek dan foreground lebih sempurna



Referensi:
Otsu, Nobuyuki, 1979, “A Threshold Selection Method from Gray-Level Histograms” IEEE Vol.9

No.1 Hal.62-68.

Comments

Popular posts from this blog

New GoPro Hero Plus, There's WiFi and Bluetooth

GoPro camera line has earned refreshment. Company known for its action camera (cam action) has released the entry-level model called Hero +. Hero + is positioned between the previous series, the Hero and Hero + LCD. That is, he has the ability on Hero and minus a few features of the Hero + LCD. Hero + itself has a body with exactly the same design with Hero. The difference, the new product is equipped with WiFi and Bluetooth wireless capabilities. WiFi and Bluetooth in Hero + can be used to access applications on smartphones GoPro. Through the application, users can control the camera, view the results of the captured video, and edit videos directly from the smartphone. the ability of the Hero + made the same as the Hero + LCD, the record 1080p video at 60fps, compared to 30 fps in Hero. The camera also has an 8-megapixel image sensor, up from 5 megapixels in Hero. Ability to dive into the water also reaches 40 meters. The difference with the Hero + LCD, Hero + is not equipped wit...

GIGS, Jadwal Konser Shaggydog

11 Oktober 2015 Kuta Beach, Kuta, Bali 4 Oktober 2015 Single Fin, Uluwatu, Bali 3 Oktober 2015 Bali Wake Park, Benoa, Bali 1 Oktober 2015 The cashew tree, Bingin, Bali 30 September 2015 Lapangan Kala hitam, Subang

Describe feelings, Facebook Presents Emoji Features such as Path

Facebook does not present a Dislike button features, but social networking is confirmed it will bring new features to Like. In addition to drawing the thumb pointed up that has become an icon Like Facebook, users can now express their liking with emoji that express love, laughter, mischievous smile, shock, sadness and anger. The feature is currently being tested in Facebook users in regions of Ireland and Spain, and will soon be rolled to other users around the world. Last month, Facebook CEO Mark Zuckerberg has indicated the presence of new features by saying that users will be able to express emotions more broadly on Facebook.