- 1_INITIAL RoI-MEAN
- The initial mean value from a circular RoI is maintained for the following horizontally shifted positions.
This is accomplished by vertically searching for the mean value that comes closest to the initial one.
- The method is fast and tolerant against noise as well as irregular edges.
- The method fails for edges of varying intensity or contrast.
-
- 2_BALANCED RoI-HISTOGRAMS
- For every horizontal position of a circular RoI, the histogram of its values is computed.
The vertical positions are those, for which the individusl histogram is balanced.
- The method is non-parametric and tolerant against intensity as well as contrast fluctuations.
- The method fails for irregular and especially interrupted edges.
-
- EXAMPLE
- Test images (8 bit):
- (a) Model edge of contrast 0.5
- (b) Gauss-filtered edge (profile length 85px)
- (c) Added lowpass-filtered noise (max values ±32)
- (e) Same as (b) but with linearly decreasing intensity
- ( f ) Same as (e) but with added lowpass-filtered noise which means decreasing signal/noise-ratio
- Edge traces:
- (b) Method_1_ (can also be obtained by proper global thresholding); Method_2_;
- (c) Method_1_ (cannot be obtained by pure global thresholding); Method_2_;
- (d) Traces from (c) super-imposed onto (a)
- (e) Method_1_ (failure due to decreasing intensity); Method_2_;
- ( f ) Method_1_ (failure due to decreasing intensity); Method_2_;
- (g) Traces from (f) super-imposed onto (a) (intensity gradient hardly affects Method_2_ result)