connected filters
play

Connected Filters Alexandre Xavier Falc ao Instituto de Computa c - PowerPoint PPT Presentation

Connected Filters Alexandre Xavier Falc ao Instituto de Computa c ao - UNICAMP afalcao@ic.unicamp.br Alexandre Xavier Falc ao MO443/MC920 - Introdu c ao ao Proc. de Imagem Digital Introduction Mathematical morphology offers


  1. Superior reconstruction The superior reconstruction of ˆ I from ˆ V 0 requires V 0 ( t ) ≥ I ( t ) for all t ∈ D I . Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  2. Superior reconstruction The superior reconstruction of ˆ I from ˆ V 0 requires V 0 ( t ) ≥ I ( t ) for all t ∈ D I . It repeats Ψ E ( ˆ V 0 , A 1 ) ∪ ˆ I multiple times up to the idempotence: Ψ E (Ψ E ( ˆ V 0 , A 1 ) ∪ ˆ I , A 1 ) ∪ ˆ I . . . ) Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  3. Superior reconstruction by IFT Instead of that, for every point t , the IFT finds a path from a regional minimum in ˆ V 0 (component X ) whose maximum altitude to reach t along that path is minimum. X 1 0 1 1 ˆ ˆ ˆ I = ( D I , I ) V 0 = ( D I , V 0 ) V = ( D I , V ) Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  4. Superior reconstruction by IFT The IFT minimizes V ( t ) = ∀ π t ∈ Π( D I , A 1 , t ) { f srec ( π t ) } min where f srec is defined by f srec ( � t � ) = V 0 ( t ) f srec ( π s · � s , t � ) = max { f srec ( π s ) , I ( t ) } . Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  5. Superior reconstruction by IFT Indeed, the problem could also be easily solved without the closing operation, by marker imposition � I ( t ) if t ∈ S , V 0 ( t ) = + ∞ otherwise, where S represents seed spels (e.g., the border of ˆ I ). Original image of a carcinoma. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  6. Superior reconstruction by IFT Indeed, the problem could also be easily solved without the closing operation, by marker imposition � I ( t ) if t ∈ S , V 0 ( t ) = + ∞ otherwise, where S represents seed spels (e.g., the border of ˆ I ). Original image of a carcinoma. Its binarization. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  7. Superior reconstruction by IFT Indeed, the problem could also be easily solved without the closing operation, by marker imposition � I ( t ) if t ∈ S , V 0 ( t ) = + ∞ otherwise, where S represents seed spels (e.g., the border of ˆ I ). Original image of a carcinoma. Its binarization. A closing of basins (marker imposition). Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  8. Superior reconstruction by IFT Indeed, the problem could also be easily solved without the closing operation, by marker imposition � I ( t ) if t ∈ S , V 0 ( t ) = + ∞ otherwise, where S represents seed spels (e.g., the border of ˆ I ). Original image of a carcinoma. Its binarization. A closing of basins (marker imposition). Its residue. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  9. Superior reconstruction by IFT Indeed, the problem could also be easily solved without the closing operation, by marker imposition � I ( t ) if t ∈ S , V 0 ( t ) = + ∞ otherwise, where S represents seed spels (e.g., the border of ˆ I ). Original image of a carcinoma. Its binarization. A closing of basins (marker imposition). Its residue. An opening by reconstruction. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  10. Inferior reconstruction by IFT Similarly, the inferior reconstruction of ˆ I from ˆ V 0 requires ≤ V 0 ( t ) I ( t ) for all t ∈ D I in order to eliminate domes rather than basins. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  11. Inferior reconstruction by IFT Similarly, the inferior reconstruction of ˆ I from ˆ V 0 requires ≤ V 0 ( t ) I ( t ) for all t ∈ D I in order to eliminate domes rather than basins. In this case, for every point t , the IFT finds a path from a regional maxima in ˆ V 0 whose minimum altitude to reach t along that path is maximum. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  12. Inferior reconstruction by IFT The IFT maximizes V ( t ) = ∀ π t ∈ Π( D I , A 1 , t ) { f irec ( π t ) } max for path function f irec defined by f irec ( � t � ) = V 0 ( t ) f irec ( π s · � s , t � ) min { f irec ( π s ) , I ( t ) } . = Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  13. Inferior reconstruction by IFT The IFT maximizes V ( t ) = ∀ π t ∈ Π( D I , A 1 , t ) { f irec ( π t ) } max for path function f irec defined by f irec ( � t � ) = V 0 ( t ) f irec ( π s · � s , t � ) min { f irec ( π s ) , I ( t ) } . = Marker imposition using a set S of seed spels is also valid. � I ( t ) if t ∈ S , V 0 ( t ) = −∞ otherwise. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  14. Superior and inferior reconstructions Therefore, we define the superior reconstruction by Ψ srec (ˆ I , ˆ V 0 , A 1 ) , ˆ V 0 ≥ ˆ I , Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  15. Superior and inferior reconstructions Therefore, we define the superior reconstruction by Ψ srec (ˆ I , ˆ V 0 , A 1 ) , ˆ V 0 ≥ ˆ I , the inferior reconstruction by Ψ irec (ˆ I , ˆ V 0 , A 1 ) , ˆ V 0 ≤ ˆ I . Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  16. Superior and inferior reconstructions Therefore, we define the superior reconstruction by Ψ srec (ˆ I , ˆ V 0 , A 1 ) , ˆ V 0 ≥ ˆ I , the inferior reconstruction by Ψ irec (ˆ I , ˆ V 0 , A 1 ) , ˆ V 0 ≤ ˆ I . The way ˆ V 0 is created gives other specific names to them. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  17. Superior and inferior reconstructions For instance, Closing by reconstruction: ˆ V 0 = Ψ C (ˆ I , A r ). Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  18. Superior and inferior reconstructions For instance, Closing by reconstruction: ˆ V 0 = Ψ C (ˆ I , A r ). Opening by reconstruction: ˆ V 0 = Ψ O (ˆ I , A r ). Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  19. Superior and inferior reconstructions For instance, Closing by reconstruction: ˆ V 0 = Ψ C (ˆ I , A r ). Opening by reconstruction: ˆ V 0 = Ψ O (ˆ I , A r ). h -Basins: residue Ψ srec (ˆ I , ˆ V 0 ) − ˆ I , ˆ V 0 = ˆ I + h , and h ≥ 1. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  20. Superior and inferior reconstructions For instance, Closing by reconstruction: ˆ V 0 = Ψ C (ˆ I , A r ). Opening by reconstruction: ˆ V 0 = Ψ O (ˆ I , A r ). h -Basins: residue Ψ srec (ˆ I , ˆ V 0 ) − ˆ I , ˆ V 0 = ˆ I + h , and h ≥ 1. h -domes: residue ˆ I − Ψ irec (ˆ I , ˆ V 0 ), ˆ V 0 = ˆ I − h , and h ≥ 1. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  21. Superior and inferior reconstructions For instance, Closing by reconstruction: ˆ V 0 = Ψ C (ˆ I , A r ). Opening by reconstruction: ˆ V 0 = Ψ O (ˆ I , A r ). h -Basins: residue Ψ srec (ˆ I , ˆ V 0 ) − ˆ I , ˆ V 0 = ˆ I + h , and h ≥ 1. h -domes: residue ˆ I − Ψ irec (ˆ I , ˆ V 0 ), ˆ V 0 = ˆ I − h , and h ≥ 1. Closing of basins or opening of domes: ˆ V 0 is created by marker imposition. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  22. Levelings Superior and inferior reconstructions can also be combined into a leveling transformation to correct edge blurring created by linear smoothing [6]. Original image. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  23. Levelings Superior and inferior reconstructions can also be combined into a leveling transformation to correct edge blurring created by linear smoothing [6]. Original image. Regular Gaussian filtering. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  24. Levelings Superior and inferior reconstructions can also be combined into a leveling transformation to correct edge blurring created by linear smoothing [6]. Original image. Regular Gaussian filtering. Leveling transformation. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  25. Levelings This leveling operator uses the following sequence of transformations from ˆ I and the impaired image ˆ V 0 . Algorithm – Leveling algorithm X ← Ψ D ( ˆ V 0 , A 1 ) ∩ ˆ 1. I. I R ← Ψ iref (ˆ 2. I , X , A 1 ) . Y ← Ψ E (ˆ 3. I , A 1 ) ∪ I R . 4. S R ← Ψ srec ( I R , Y , A 1 ) . Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  26. Superior reconstruction computation For superior reconstruction: First, all nodes t ∈ D I are trivial paths with initial connectivity values V 0 ( t ). Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  27. Superior reconstruction computation For superior reconstruction: First, all nodes t ∈ D I are trivial paths with initial connectivity values V 0 ( t ). The initial roots are identified at the global minima of V 0 ( t ). Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  28. Superior reconstruction computation For superior reconstruction: First, all nodes t ∈ D I are trivial paths with initial connectivity values V 0 ( t ). The initial roots are identified at the global minima of V 0 ( t ). They may conquer their adjacent nodes by offering them better paths. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  29. Superior reconstruction computation For superior reconstruction: First, all nodes t ∈ D I are trivial paths with initial connectivity values V 0 ( t ). The initial roots are identified at the global minima of V 0 ( t ). They may conquer their adjacent nodes by offering them better paths. The process continues from the adjacent nodes in a non-decreasing order of path values. if max { f srec ( π s ) , I ( t ) } < f srec ( π t ) then π t ← π s · � s , t � . Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  30. Superior reconstruction computation For superior reconstruction: First, all nodes t ∈ D I are trivial paths with initial connectivity values V 0 ( t ). The initial roots are identified at the global minima of V 0 ( t ). They may conquer their adjacent nodes by offering them better paths. The process continues from the adjacent nodes in a non-decreasing order of path values. if max { f srec ( π s ) , I ( t ) } < f srec ( π t ) then π t ← π s · � s , t � . Essentially the regional minima in V 0 ( t ) compete among themselves and some of them become roots (i.e., minima in V ( t )). Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  31. Superior reconstruction computation The optimum-path forest with filtered values V ( t ) (right) resulting from the superior reconstruction of ˆ I = ( D I , I ) (left) from marker ˆ V 0 = ( D I , V 0 ) (center) contains unconquered regions (black dots) and the winner regional minima (red dots) as roots. 8 �� �� 30 8 25 10 10 3 8 25 � � �������� �������� ��� ��� � � ��� ��� � � 20 � � ��� ��� 5 20 � � 8 15 15 � � 25 20 ��� ��� 20 30 � � 30 40 ��� ��� 5 5 0 � � ����� ����� 10 15 10 20 20 20 �� �� � � ����� ����� 8 8 �� �� �� �� ����� ����� �� �� 4 10 5 10 12 10 5 ����� ����� Images ˆ I (left), ˆ V 0 (center), and ˆ V (right). Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  32. Superior reconstruction algorithm Algorithm – Superior reconstruction algorithm 1. For each t ∈ D I , do 2. Set V ( t ) ← V 0 ( t ) . 3. If V ( t ) � = + ∞ , then insert t in Q. 4. While Q is not empty, do 5. Remove from Q a spel s such that V ( s ) is minimum. For each t ∈ A 1 ( s ) such that V ( t ) > V ( s ) , do 6. Compute tmp ← max { V ( s ) , I ( t ) } . 7. 8. If tmp < V ( t ) , then If V ( t ) � = + ∞ , remove t from Q. 9. Set V ( t ) ← tmp. 10. 11. Insert t in Q. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  33. Organization of this lecture Basic definitions. Superior and inferior reconstructions. Their relation with watershed-based segmentation. Fast binary filtering. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  34. Superior reconstruction and watershed transform Suppose we make a hole in each minimum of an image ˆ I and submerge its surface in a lake, such that each hole starts a flooding with water of different color. A watershed segmentation is obtained by preventing the mix of water from different colors. Original image ˆ I . Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  35. Superior reconstruction and watershed transform Suppose we make a hole in each minimum of an image ˆ I and submerge its surface in a lake, such that each hole starts a flooding with water of different color. A watershed segmentation is obtained by preventing the mix of water from different colors. Original image ˆ I . IFT-watershed segmentation. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  36. Superior reconstruction and watershed transform Suppose we make a hole in each minimum of an image ˆ I and submerge its surface in a lake, such that each hole starts a flooding with water of different color. A watershed segmentation is obtained by preventing the mix of water from different colors. Original image ˆ I . IFT-watershed segmentation. Classical watershed segmentation requires to detect and label each minimum before the flooding process. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  37. Superior reconstruction and watershed transform During superior reconstruction, we may force each regional minimum in ˆ I to produce a single optimum-path tree in P with a distinct label in L . Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  38. Superior reconstruction and watershed transform During superior reconstruction, we may force each regional minimum in ˆ I to produce a single optimum-path tree in P with a distinct label in L . By definition, the resulting optimum-path forest is a watershed segmentation. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  39. Superior reconstruction and watershed transform During superior reconstruction, we may force each regional minimum in ˆ I to produce a single optimum-path tree in P with a distinct label in L . By definition, the resulting optimum-path forest is a watershed segmentation. Moreover, by choice of ˆ V 0 , we may also eliminate the influence zones of “irrelevant” minima and considerably reduce the over-segmentation problem. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  40. Superior reconstruction and watershed transform During superior reconstruction, we may force each regional minimum in ˆ I to produce a single optimum-path tree in P with a distinct label in L . By definition, the resulting optimum-path forest is a watershed segmentation. Moreover, by choice of ˆ V 0 , we may also eliminate the influence zones of “irrelevant” minima and considerably reduce the over-segmentation problem. A change of topology in Ψ srec (ˆ I , ˆ V 0 , A r ) for r > 1 also helps on that. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  41. Superior reconstruction and watershed transform This requires a simple modification in f srec . � I ( t ) if t ∈ R , f srec ( � t � ) = V 0 ( t ) + 1 otherwise, f srec ( π s · � s , t � ) = max { f srec ( π s ) , I ( t ) } , where R is found on-the-fly with a single root for each regional minimum of the filtered image ˆ V . The condition V 0 ( t ) + 1 > I ( t ) guarantees that all spels in D I will be conquered. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  42. Superior reconstruction and watershed transform The choice of V 0 ( t ) = I ( t ) + h , h ≥ 0 will preserve all minima of ˆ I whose basins have depth greater than h . For h = 0, all minima will be preserved. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  43. Superior reconstruction and watershed transform The choice of V 0 ( t ) = I ( t ) + h , h ≥ 0 will preserve all minima of ˆ I whose basins have depth greater than h . For h = 0, all minima will be preserved. 20 18 20 21 23 20 5 8 5 10 13 10 (b) (c) (a) (a) Image ˆ I . (b) Image ˆ V 0 + 1 for h = 2. (c) Image V = Ψ srec (ˆ ˆ I , ˆ V 0 , A 1 ) with indication of optimum paths in P . Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  44. Superior reconstruction and watershed transform The choice of V 0 ( t ) = I ( t ) + h , h ≥ 0 will preserve all minima of ˆ I whose basins have depth greater than h . For h = 0, all minima will be preserved. 18 18 20 19 21 20 20 5 6 5 10 11 10 (b) (c) (a) (a) Image ˆ I . (b) Image ˆ V 0 + 1 for h = 0. (c) Image V = Ψ srec (ˆ ˆ I , ˆ V 0 , A 1 ) with indication of optimum paths in P . Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  45. Watershed from grayscale marker For grayscale images ˆ V 0 , the simultaneous computation of a superior reconstruction in ˆ V and a watershed segmentation in L is called watershed from grayscale marker [4]. MR-image of a wrist. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  46. Watershed from grayscale marker For grayscale images ˆ V 0 , the simultaneous computation of a superior reconstruction in ˆ V and a watershed segmentation in L is called watershed from grayscale marker [4]. MR-image of a wrist. A gradient image ˆ I . Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  47. Watershed from grayscale marker For grayscale images ˆ V 0 , the simultaneous computation of a superior reconstruction in ˆ V and a watershed segmentation in L is called watershed from grayscale marker [4]. MR-image of a wrist. A gradient image ˆ I . The closing V 0 = Ψ C (ˆ ˆ I , A 2 . 5 ). Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  48. Watershed from grayscale marker For grayscale images ˆ V 0 , the simultaneous computation of a superior reconstruction in ˆ V and a watershed segmentation in L is called watershed from grayscale marker [4]. MR-image of a wrist. A gradient image ˆ I . The closing V 0 = Ψ C (ˆ ˆ I , A 2 . 5 ). Segmentation in L for Ψ srec (ˆ I , ˆ V 0 , A 3 . 5 ). Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  49. Watershed from grayscale marker Algorithm – Watershed from Grayscale Marker 1. For each t ∈ D I , do 2. Set P ( t ) ← nil, λ ← 1 , and V ( t ) ← V 0 ( t ) + 1 . 3. Insert t in Q. 4. While Q is not empty, do 5. Remove from Q a spel s such that V ( s ) is minimum. If P ( s ) = nil then set V ( s ) ← I ( s ) , L ( s ) ← λ , and λ ← λ + 1 . 6. For each t ∈ A ( s ) such that V ( t ) > V ( s ) , do 7. Compute tmp ← max { V ( s ) , I ( t ) } . 8. 9. If tmp < V ( t ) , then Set P ( t ) ← s, V ( t ) ← tmp, L ( t ) ← L ( s ) . 10. 11. Update position of t in Q. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  50. Organization of this lecture Basic definitions. Superior and inferior reconstructions. Their relation with watershed-based segmentation. Fast binary filtering. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  51. Fast binary filtering via IFT For binary images ˆ I and Euclidean relations A r , it is also possible to exploit the IFT for fast computation of morphological operators, which can be decomposed into alternate sequences of erosions and dilations (or vice-versa). For instance, Ψ C (ˆ Ψ E (Ψ D (ˆ I , A r ) = I , A r ) , A r ) . Ψ CO (ˆ Ψ D (Ψ E (Ψ E (Ψ D (ˆ I , A r ) = I , A r ) , A r ) , A r ) , A r ) Ψ D (Ψ E (Ψ D (ˆ = I , A r ) , A 2 r ) , A r ) . Ψ CO (Ψ CO (ˆ Ψ D (Ψ E (Ψ D (Ψ E (Ψ D (ˆ I , A r ) , A 2 r ) I , A r ) , A 2 r ) , = A 3 r ) , A 4 r ) , A 2 r ) . Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  52. Fast binary filtering via IFT The basic idea is to extract the object’s (background’s) border S , Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  53. Fast binary filtering via IFT The basic idea is to extract the object’s (background’s) border S , compute their propagation in sub-linear time outward (inward) the object for dilation (erosion), alternately. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  54. Fast binary filtering via IFT The basic idea is to extract the object’s (background’s) border S , compute their propagation in sub-linear time outward (inward) the object for dilation (erosion), alternately. Each border propagation stops at the adjacency radius specified for dilation (erosion). Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  55. Fast binary filtering via IFT This requires to constrain the computation of an Euclidean distance transform (EDT) either outside (dilation) or inside (erosion) the object up to a distance r from it. The EDT assigns to every spel in D I its distance to the closest spel in a given set S ⊂ D I (e.g., the object’s or background’s border). Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  56. Fast binary filtering via IFT This requires to constrain the computation of an Euclidean distance transform (EDT) either outside (dilation) or inside (erosion) the object up to a distance r from it. r The EDT assigns to every spel in D I its distance to the closest spel in a given set S ⊂ D I (e.g., the object’s or background’s border). Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  57. Fast binary filtering via IFT A spel s ∈ D I belongs to an object’s border S , when I ( s ) = 1 and ∃ t ∈ A 1 ( s ), such that I ( t ) = 0. Similar definition applies to backgroud’s border. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  58. Fast binary filtering via IFT A spel s ∈ D I belongs to an object’s border S , when I ( s ) = 1 and ∃ t ∈ A 1 ( s ), such that I ( t ) = 0. Similar definition applies to backgroud’s border. For dilation, the value 1 is propagated to every spel t with value I ( t ) = 0 and distance � t − R ( π t ) � 2 ≤ r 2 , R ( π t ) ∈ S . Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  59. Fast binary filtering via IFT A spel s ∈ D I belongs to an object’s border S , when I ( s ) = 1 and ∃ t ∈ A 1 ( s ), such that I ( t ) = 0. Similar definition applies to backgroud’s border. For dilation, the value 1 is propagated to every spel t with value I ( t ) = 0 and distance � t − R ( π t ) � 2 ≤ r 2 , R ( π t ) ∈ S . For erosion, the value 0 is propagated to every spel t with value I ( t ) = 1 and distance � t − R ( π t ) � 2 ≤ r 2 , R ( π t ) ∈ S . Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  60. Fast binary filtering via IFT A spel s ∈ D I belongs to an object’s border S , when I ( s ) = 1 and ∃ t ∈ A 1 ( s ), such that I ( t ) = 0. Similar definition applies to backgroud’s border. For dilation, the value 1 is propagated to every spel t with value I ( t ) = 0 and distance � t − R ( π t ) � 2 ≤ r 2 , R ( π t ) ∈ S . For erosion, the value 0 is propagated to every spel t with value I ( t ) = 1 and distance � t − R ( π t ) � 2 ≤ r 2 , R ( π t ) ∈ S . During dilation (erosion), spels t whose distance � t − R ( π t ) � 2 > r 2 but � P ( t ) − R ( π t ) � 2 ≤ r 2 are stored in a new set S ′ for a subsequent erosion (dilation) operation. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  61. Fast binary filtering via IFT The EDT is propagated in V from a set S ⊂ D I to every spel t ∈ D I in a non-decreasing order of squared distance using A √ 2 in 2D (8-neighbors) [7]. For fast dilation, it uses path function  0 if t ∈ S ,  f euc ( � t � ) + ∞ = if I ( t ) = 0, −∞ otherwise.  � t − R ( π s ) � 2 . f euc ( π s · � s , t � ) = Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  62. Fast binary filtering via IFT For fast erosion, it uses path function  0 if t ∈ S ,  f euc ( � t � ) = + ∞ if I ( t ) = 1, −∞ otherwise.  � t − R ( π s ) � 2 . f euc ( π s · � s , t � ) = A dilated (eroded) binary image J = ( D I , J ) is created during the distance propagation process. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  63. Fast dilation Algorithm – Fast Dilation in 2D up to distance r from S 1. For each t ∈ D I , set J ( t ) ← I ( t ) , R ( π t ) ← t and V ( t ) ← f euc ( � t � ) . 2. While S � = ∅ , remove t from S and insert t in Q. 3. While Q is not empty, do 4. Remove from Q a spel s such that V ( s ) is minimum. if V ( s ) ≤ r 2 , then 5. Set J ( t ) ← 1 . 6. For each t ∈ A √ 7. 2 ( s ) such that V ( t ) > V ( s ) , do Compute tmp ← � t − R ( π s ) � 2 . 8. 9. If tmp < V ( t ) , then 10. If V ( t ) � = + ∞ , remove t from Q. Set V ( t ) ← tmp and R ( π t ) ← R ( π s ) . 11. 12. Insert t in Q. Else insert s in S . 13. Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  64. Fast binary filtering via IFT Sets S and S ′ may contain spels from multiple borders. Multiple borders, Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  65. Fast binary filtering via IFT Sets S and S ′ may contain spels from multiple borders. Multiple borders, distances outside up to r = 10, Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  66. Fast binary filtering via IFT Sets S and S ′ may contain spels from multiple borders. Multiple borders, distances outside up to r = 10, their dilation, Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  67. Fast binary filtering via IFT Sets S and S ′ may contain spels from multiple borders. Multiple borders, distances outside up to r = 10, their dilation, erosion, Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  68. Fast binary filtering via IFT Sets S and S ′ may contain spels from multiple borders. Multiple borders, distances outside up to r = 10, their dilation, erosion, closing, Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  69. Fast binary filtering via IFT Sets S and S ′ may contain spels from multiple borders. Multiple borders, distances outside up to r = 10, their dilation, erosion, closing, closing by reconstruction, Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

  70. Fast binary filtering via IFT Sets S and S ′ may contain spels from multiple borders. Multiple borders, distances outside up to r = 10, their dilation, erosion, closing, closing by reconstruction, opening, and Alexandre Xavier Falc˜ ao MO443/MC920 - Introdu¸ c˜ ao ao Proc. de Imagem Digital

Recommend


More recommend