Table 1 Filtering operations and the filter parameters for computing pixel-level features from training images.
From: Multi-scale Gaussian representation and outline-learning based cell image segmentation
Operation (Feature) | Parameter | Values | Total |
|---|---|---|---|
Gaussian low pass | kernel width σ | 3:2:49 | 24 |
Integrated pixel intensity | kernel size | 3:2:9 | 04 |
Laplacian of Gaussian | kernel width σ | 3:2:49 | 24 |
Difference of Gaussian | kernel width σ | 05 | |
Morphological top-hat | kernel size | 3:2:49 | 24 |
Morphological bottom-hat | kernel size | 3:2:49 | 24 |
Local binary pattern | (quantization, | ||
and contrast | radius) | (8,1) | 02 |
Variance | kernel size | 3:2:49 | 24 |
Order statistics | |||
(Min., Med., Max.) | kernel size | 3:2:7 | 09 |
Haralick (13-features) | kernel size | 5:2:15 | 78 |
Gabor filter | kernel size, | 5:2:15, | |
freq. f, | 1/4:1/4:3/4, | ||
orientation θ | 0:π/4:3π/4 | 72 | |
Total number of features. | 290 |