Page History
...
Module ImgAlgos::CSPadArrPeakFinder
Wiki Markup |
---|
ThisIn this module the "median algorithm" is intendedimplemented for peak finding in CSPad data array shaped as \[5920=4*8*185\]\[388\]. It extends features of the module [ImgAlgos::CSPadArrNoise|PCDS:Psana Module Catalog#Module ImgAlgos::CSPadArrNoise]) and works as follows\]. The "median algorithm" assumes that the amplitude level of background and noise for each pixel can be estimated as a mean and RMS of the surrounding pixels, located in the ring with parameters {{rmin}} and {{dr}} around the pixel in question. The threshold {{SoNThr_noise}} on signal over noise (S/N) ratio allow to asset the pixel amplitude as a large noise fluctuation. Statistics of pixels above the S/N threshold accumulated over many images can be used to form the noisy-pixel mask. For example, if the fraction of images where pixel exceeds the S/N threshold grater than certain value ({{frac_noisy_imgs}}=0.9), the pixel is considered as noisy. The permanent bad pixel mask (see module [ImgAlgos::CSPadMaskApply|PCDS:Psana Module Catalog#Module ImgAlgos::CSPadMaskApply]) and dynamically evaluated noisy pixel mask are used to get rid of bad pixels and improve the image quality. Healthy pixels with S/N above threshold ({{SoNThr_signal}} about 3-5) are treated as potential signals. Using recursive flood-filling algorithm the groups of connected signal pixels can be found and considered as a candidate for a diffraction peaks. Peak finding algorithm uses the amplitude, S/N thresholds, and limits on number of pixels in the connected region (parameters {{peak_amp_tot_thr}}, {{peak_SoN_thr}}, {{peak_npix_min}}, and {{peak_npix_max}}) in order to define the peak. Finally, the event is selected or discarded depending on number of found peaks and total amplitude threshold, defined by the parameters {{event_npeak_min}}, {{event_npeak_max}}, and {{event_amp_tot_thr}}, respectively. |
Description of implemented algorithm:
- in the constructor and
beginJob(...)
method:- enters enter input parameters,
- (re)set the initial mask of noisy pixels from file
hot_pix_mask_file
(if its name is specified in the configuration file), - do necessary initialization of work arrays.
- in the
event(...)
method : uses the main part of "median algorithm" is implemented:- to evaluate the signal and noise for each pixel,
Wiki Markup fillsfill \[4\]\[8\]\[185\]\[388\] per-pixel arrays:
m_stat
- number of events with|S/N| > SoNThr
,m_signal
- signal amplitude, or 0(zero) for masked pixels,m_proc_status
- sets 255 forS/N > SoNThr
or 0(zero) for masked pixels.
- use arrays
m_proc_status
andm_signal
to find peaks:Wiki Markup iterate over \[185\]\[388\] 2x1 pixels and find the connected regions (using recursive flood-filling algorithm)
- create vector of peaks
v_peaks
ofstruct Peak
, usingpeak_npix_min
,peak_npix_max
, andpeak_amp_tot_thr
parameters,
- loop over
v_peaks
, count total amplitude and the number of peaks in the event. - decide if the event selected or not based on
event_npeak_min
,event_amp_tot_thr
, andselection_mode
parameters. - periodically dynamically re-generate the mask, based on
m_stat
array andfrac_noisy_imgs
parameter. When to start and for how many events to update the mask is defined by thenevents_mask_update
andnevents_mask_accum
parameters, respectively. - save
m_signal
in file for selected events, depending onout_file_bits
parameter. - put the vector with peaks
v_peaks
in the evt with key=key_peaks_out
.
- in the
endJob(...)
method, depending on bit status inout_file_bits
:- save current hot-pixel mask in the file
hot_pix_mask_out_file
- save current fraction of events with noisy/signal pixels in the file
frac_noisy_evts_file
- save current hot-pixel mask in the file
...
Overview
Content Tools