#include <gevd_noise.h>
Definition at line 33 of file gevd_noise.h.
Public Member Functions | |
| gevd_noise (const float *data, const int n, const int number_of_bins=200) | |
| Generate the histogram curve at low responses to estimate the sensor/texture noise in data. | |
| ~gevd_noise () | |
| Free allocated space. | |
| bool | EstimateSensorTexture (float &sensor, float &texture) const |
| Fit a Raleigh distribution to the histogram curve of edgels with low magnitudes, h(x), to estimate the sensor noise, as would be zero-crossing of dh(x), and texture noise as the dominant peak in h(x). | |
| const float * | Histogram (int &n) const |
| float | BinSize () const |
Static Public Member Functions | |
| static float * | EdgelsInCenteredROI (const gevd_bufferxy &magnitude, const gevd_bufferxy &dirx, const gevd_bufferxy &diry, int &nedgel, const int roiArea=250 *250) |
| Collect all edgels above zero, in an ROI at center of image. | |
Static Protected Member Functions | |
| static bool | WouldBeZeroCrossing (const float *dhist, const int nbin, float &index) |
| Find would be zero-crossing of the derivative of the histogram from its downward curvature. | |
| static bool | RealZeroCrossing (const float *dhist, const int nbin, float &index) |
| Find real zero-crossing of the derivative of the histogram. | |
Protected Attributes | |
| float * | hist |
| int | nbin |
| float | binsize |
Friends | |
| class | DetectionUI |
|
||||||||||||||||
|
Generate the histogram curve at low responses to estimate the sensor/texture noise in data.
Definition at line 15 of file gevd_noise.cxx. |
|
|
Free allocated space.
Definition at line 78 of file gevd_noise.cxx. |
|
|
Definition at line 50 of file gevd_noise.h. |
|
||||||||||||||||||||||||
|
Collect all edgels above zero, in an ROI at center of image. This utility function is used to collect edgels from a response image, i.e. gradient/hessian/laplacian, or other edge response image. Definition at line 89 of file gevd_noise.cxx. |
|
||||||||||||
|
Fit a Raleigh distribution to the histogram curve of edgels with low magnitudes, h(x), to estimate the sensor noise, as would be zero-crossing of dh(x), and texture noise as the dominant peak in h(x). Setting the threshold at 3 times the sensor/texture noise would eliminate 99% of all noisy edges. The raw noise in the original image can be deduced from the filter used to convolve with the image. H. Voorhees & T. Poggio, Detecting Blobs as Textons in Natural Images, Proc. 1987 IU Workshop, Los Angeles, CA. Definition at line 171 of file gevd_noise.cxx. |
|
|
Definition at line 49 of file gevd_noise.h. |
|
||||||||||||||||
|
Find real zero-crossing of the derivative of the histogram. This is texture noise in the ROI. Protected. Definition at line 269 of file gevd_noise.cxx. |
|
||||||||||||||||
|
Find would be zero-crossing of the derivative of the histogram from its downward curvature. This is sensor noise in the ROI. Protected. Definition at line 216 of file gevd_noise.cxx. |
|
|
Definition at line 36 of file gevd_noise.h. |
|
|
Definition at line 55 of file gevd_noise.h. |
|
|
Definition at line 53 of file gevd_noise.h. |
|
|
Definition at line 54 of file gevd_noise.h. |
1.4.4