Publier

ARCH-H-101 Histoire de l'architecture : le 20ème siècle

< Retour

Oral - 11 Jan 2022

Theory:
Primary question:
• Explain everything you know about morphomathematics.
— Operations on binary images.
— Erosion, dilation, opening and closing.
— Hit-or-miss.
— Practical applications (OCR, Pruning).
Follow-up questions:
• What operation allows separation of non-contiguous objects?
— Labelling.
• What operation turns a binary image into a grayscale image?
— Distance transform.

Practical:
Primary question:
• Explain the pipeline used to segment brain tumors.
— Manual, optimal and otsu thresholding.
— Manually and automatically-placed markers.
— Split-and-merge watershed transform.
— Auto-level and equalization.
Follow-up questions:
• What does the otsu threshold suppose?
— A histogram with 2 bell curves or distribution concentrated in 2 regions.
• How do you use otsu in order to segment the tumor (histogram with 3 bell curves) ?
— Use multiotsu thresholding in order to define 3 classes or you disregard the black pixels (background) and end up with 2 bell curves (one representing the brain and the other, the tumors).

Oral Jan 11th, 2022 - 11 Jan 2022

Theory:
Primary question:
• Explain everything you know about morphomathematics.
— Operations on binary images.
— Erosion, dilation, opening and closing.
— Hit-or-miss.
— Practical applications (OCR, Pruning).
Follow-up questions:
• What operation allows separation of non-contiguous objects?
— Labelling.
• What operation turns a binary image into a grayscale image?
— Distance transform.

Practical:
Primary question:
• Explain the pipeline used to segment brain tumors.
— Manual, optimal and otsu thresholding.
— Manually and automatically-placed markers.
— Split-and-merge watershed transform.
— Auto-level and equalization.
Follow-up questions:
• What does the otsu threshold suppose?
— A histogram with 2 bell curves or distribution concentrated in 2 regions.
• How do you use otsu in order to segment the tumor (histogram with 3 bell curves) ?
— Use multiotsu thresholding in order to define 3 classes or you disregard the black pixels (background) and end up with 2 bell curves (one representing the brain and the other, the tumors).


Il n'y a pas de publications plus anciennes.