Morphological Segmentation for Image Processing and Visualization

Morphological Segmentation for Image Processing and Visualization

Morphological Segmentation for Image Processing and Visualization

J.Robarts Research Institute

London,Canada

Lixu Gu

Road Map

Mathematical Morphology Image Processing:

– 2D application: Character Extraction– 3D application: Medical Image Processing

Image Visualization: BrainView– Registration and Visualization– Segmentation and Visualization

Future Works:

Mathematical Morphology

Mathematical morphology is a powerful methodology which was initiated in the late 1960s by G.Matheron and J.Serra at the Fontainebleau School of Mines in France.

nowadays it offers many theoretic and algorithmic tools inspiring the development of research in the fields of signal processing, image processing, machine vision, and pattern recognition.

Morphological Operations -1 The four most basic operations in mathematical

morphology are dilation, erosion, opening and Closing:

Dilation Erosion

Opening Closing

Morphological Operations -2

Top-hat Transformation (TT):– An excellent tool for extracting bright or An excellent tool for extracting bright or

dark objectsdark objects– cannot deal with many complicated cannot deal with many complicated

problemsproblems– Difficult to determine proper size of Difficult to determine proper size of

structuring elements automaticallystructuring elements automatically

Differential Top-hat Transformation (DTT):

Morphological Reconstruction Conditional Dilation : a special recursive dilation

operation (region growing); a powerful function to restore destroyed objective regions.

– Let M and V (M V) be two binary images defined as “marker” and “mask”, respectively.

– Conditional dilation Ri(M,V) is defined as:

– Marker M is only allowed to grow in the region restricted by mask V.

1

( , ) ( ) ,

( , ) ( , )

i

i

i i

R M V M K V

R M V R M Vuntil

Morphological Reconstruction Algorithm for binary reconstruction:

Original (V) Opened (M) Reconstructed (T)

1. M = V o K , where K is any SE.

2. T = M,

3. M= M Ki , where i=4 or i=8,

4. M = M∩ V, [Take only those pixels from M that are also in V .]

5. if M T then go to 2,

6. else stop;

1. M = V o K , where K is any SE.

2. T = M,

3. M= M Ki , where i=4 or i=8,

4. M = M∩ V, [Take only those pixels from M that are also in V .]

5. if M T then go to 2,

6. else stop;

Application in 2D Image Processing Character Extraction-1

Character Extraction From Cover Image (Source)

Application in 2D Image Processing Character Extraction-2

Character Extraction From Cover Image (Results)

Application in 2D Image Processing Character Extraction-3

MorningMorning NoonNoon

AfternoonAfternoon EveningEvening

Application in 2D Image Processing Character Extraction-4

MorningMorning NoonNoon

AfternoonAfternoon EveningEvening

Application in 3D Image Processing Organs Extraction-1

slice20slice20 slice30slice30slice25slice25

slice30slice30slice25slice25slice20slice20

Application in 3D Image Processing Organs Extraction-2

Top View Back View

Application in 3D Image Processing Organs Extraction-3

Application in 3D Image Processing Organs Extraction-4

Segmented heart beating cycle

Application in 3D Image Processing Organs Extraction-5

Kidney with Bones Kidney with Vessels

Image Visualization – BrainView

BrainView is a software which I designed and developed at J.Robarts Research Institute, London, Ontario for her industry partner : Cedara Software.

It is designed to visualize the structures of brain and its atlases for stereotaxy surgery navigation (Image Guided Neuro-Surgery).

It is under Python, VTK environment

Main Design Issues

Ac-Pc: two anatomic landmarks located in the deep brain used to define the Patient coordinate space

PGS: a Proportional Grid System is designed to segment a brain into 12 sub-regions based on the dimension derived from Ac-Pc Setting.

PWL: a Piece-Wise Linear co-registration technique to warp brain atlases into patient brain space.

Brain View snapshot -1

PGS in a patient brain

Brain View snapshot -2

Co-registered atlas using PWL

Brain View snapshot-3--Registration tool kit

Features:1. Cut plane in 3D2. Work in 2 data sets3. 2D and 3D view4. Registration methods:

• LandMark• ThinPlateSpline• GridTransform• MutualInformation

Mutual Information Registration

Brain View snapshot-4--Segmenation tool kit

Features:1. Cut plane in 3D2. Work in 2 data sets3. 2D and 3D view4. Segmenation

methods:• Morphology• Snake• Level Set• Watershed

Research Plan-1 Medical Image Analysis

--- Segmentation and Registration

– More efforts address on Ultrasound Image (2D, 3D)

Segmented baby face from US

Real time US, MR integration for IGS2D Segmentation using GDM

Research Plan-2 Image Guided Surgery and Therapy:

– Neuro Surgical Navigation1. Patient data acquisition2. Image Visualization3. Surgical Plan4. Surgical Navigation

– Cardiac Surgical Navigation

Research Plan-3 Virtual Human:

-- Set up a virtual reality human

model for surgery plan

and navigation in the future.

Virtual Training and Planning

Research Plan-4

Robotic Surgery Navigation:--- Work on human interface

Research Plan-5 Functional MRI (fMRI) for mind study

– Research on computer aided acupuncture• Find the relationship between acupoint and other

organs using fMRI, PET or SPECT technology

• Visualize acupoint in the human body (eg. Visible Chinese)

• Find the best procedure for image-guided acupuncture

– Other mind study : Vision, Neurosurgical plan, Language, Pain, et.al.

Question