11
Automated detection of filaments on full disk H images
Nicolas Fuller and Jean Aboudarham
Meudon Observatory / LESIA
October 2003
EGSO WP5
22
EGSO
Automated Detection of Filaments / NF & JA 2003
Original imageOriginal image
StandardizationStandardization
CleaningCleaning
Seeds detectionSeeds detection
Region growingRegion growing
Shape descriptionShape description
Catalogue parametersCatalogue parameters
33
EGSO
Cleaning Process
Darkening Dust lines
Automated Detection of Filaments / NF & JA 2003
44
EGSO
Cleaning process / Intensity normalization 1
• Need to compute the slow variations of the background• Use of a large median filter on a resized image• The first approximation is influenced by large bright plages and large filaments
1. Resize the image I to a smaller scale -> Is2. Apply a median filter with a large window to Is3. Resize to original scale (-> B) and subtract from I -> I’4. From I’define 2 thresholds to roughly locate filaments and bright plages5. Replace their value with the corresponding values in B -> I”6. Apply step 1 to 3 to I” and get the final background and the normalize image
Automated Detection of Filaments / NF & JA 2003
55
EGSO
Cleaning process / Intensity normalization 2
- =
- =
Automated Detection of Filaments / NF & JA 2003
66
EGSO
Cleaning process / Dust lines removal 1
Need to compute a binary image with most of theNeed to compute a binary image with most of theline points set to 1 and the background to 0 :line points set to 1 and the background to 0 :• ThresholdThreshold• Thinning morphological operatorThinning morphological operator
Automated Detection of Filaments / NF & JA 2003
77
EGSO
Cleaning process / Dust lines removal 2
Original Threshold Thinning
Hough transform
Houghbackprojection
Threshold
Line pixelslocations
Pixels valuescorrection
Automated Detection of Filaments / NF & JA 2003
88
EGSO
Automated Detection of Filaments / NF & JA 2003
Image enhancement
To enhance the image sharpness we use a Laplacian filter• Filaments contours are better defined• Allows to detect the thinnest parts of the filaments more efficiently
g(x,y) = f(x,y) –2f(x,y) where 2f = 2f/x2 + 2f/y2
11 11 11
11 -8-8 11
11 11 11
Digital implementation:
Before and after enhancement
99
EGSO
Automated Detection of Filaments / NF & JA 2003
Region Growing
Definition: “procedure that groups pixels into larger regions based on predefined criteria. It starts with a set of ‘seed’ points and from these grow regions by appending to each seed those neighboring pixels that have properties similar to the seed”
1010
EGSO
Automated Detection of Filaments / NF & JA 2003
Region Growing / Seeds detection
To find the seed points we apply a windowed threshold:The pixels statistics in each window (200*200) are computed and the
threshold is given by: Twin = Mwin – x win
M : Mean
: constant
: standard deviation
1111
EGSO
Automated Detection of Filaments / NF & JA 2003
Region Growing
For each seed we define an intensity range which is a criteria to append connected pixels to the seed: [ 0, Tbr ]
where Tbr = Mbr – x br
( br stands for Bounding Rectangle )A minimum region size is also defined
1212
EGSO
Automated Detection of Filaments / NF & JA 2003
Region growing / BBSO
The process has been tested on other H full disk observations : Big Bear Solar Observatory example
1313
EGSO
Automated Detection of Filaments / NF & JA 2003
Shape analysis / Morphological operators
Morphological closing
Morphological thinning / pruning
Skeleton Length/centre/Chain Code…
Chain codedirection numbers
1414
EGSO
Automated Detection of Filaments / NF & JA 2003
Parameters : examples
GRAV_C_CAR_LAT DOUBLE -18.297280 GRAV_C_CAR_LON DOUBLE 337.64961 BRPIX_X_LL DOUBLE 562.00000 BRPIX_Y_LL DOUBLE 418.00000 BRPIX_X_UR DOUBLE 574.00000 BRPIX_Y_UR DOUBLE 430.00000 SAMPLECOUNT LONG 58 AREA DOUBLE 2.8334533 SKE_LEN_DEG DOUBLE 2.254509 ELONG DOUBLE 0.90625000 MEAN_INT_RATIO DOUBLE 0.83272228 FEAT_MAX_INT DOUBLE 978.00000 FEAT_MIN_INT DOUBLE 689.00000 FEAT_MEAN_INT DOUBLE 869.36206 ENC_MET STRING 'CHAINCODE' COD_PIX_X DOUBLE 572.00000 COD_PIX_Y DOUBLE 417.00000 COD_SKE_PIX_X DOUBLE 574.00000 COD_SKE_PIX_Y DOUBLE 418.00000 SKE_CHAIN STRING '33332233334343' BND_CHAIN STRING '00123322222234433443444566707777677076'
1515
EGSO
Automated Detection of Filaments / NF & JA 2003