ノート/ノート
Edge Detection (輪郭抽出) †
Visitors: 2007 Last Modified: 2008-08-11 (月) 10:59:46
Properties
- Already known to be a cell, i.e., the edges form a closed loop.
- Color and contrast are not known in advance. Detection needs to be adaptive.
- Color and contrast depend on its place, even within the same cell. This is
possibly due to the vertical position of the cell, light beam position, etc.
Possible adaptive detection
- We can ask for users to designate two positions about the cell, 1) internal
"mid-position", and 2) external position.
- The program can draw a line between these two points, and measure the
color changes along the line.
- The program can designate a "possible edge point" along the line. This should
be only one point along the line.
- If appropriate, the program asks the user to "correct" the detected edge point. I am not sure if it is possible or not to use the corrected point as the edge.
- Next, starting the established "edge point", the program moves the external
position (2 above) along the circle with the internal position as its center and the external position as its radius.
- If the edge point goes outside the circle while moving around the center, the program can detect it by the movement of the edge point along the radius, and expand the length of the radius by 5-10% or so. More radical expansion of the radius may gather edges of other cells and not desirable.
- In the case of lost tracking while circling, possibly need to skip several points. If skipping goes beyond the upper limit, then, we should restart the edge point tracking in the reverse direction from the first position. This may either fully go around the center, in which case the tracking in the original direction was wrong in some sense, or get lost at some point, in which case the entire edge line is not a closed line.
In the disconnected case, we either 1)draw an imaginaly edge, 2) try to change edge detection parameters to find a edge point, or 3) give up edge detection.
Memos
![[Toho University]](image/CLogo_Kihon_process.jpg "[Toho University]")
![[YYLAB]](image/yamanouc2.png "[YYLAB]")
