We thank the referee for providing constructive comments and help in improving the contents of this paper. According to your comments, we made the following revisions. - This paper describes the application of advanced image processing techniques, such as SIDE and SVM, to the detection of solar filaments in H-alpha images. While the methods are novel and the results impressive, the paper is poorly written and poorly organized. In particular, the use of English is not of a sufficient level for publication in Solar Physics. The authors should therefore

We thank the referee for providing constructive comments and help

in improving the contents
of this paper.

According to your comments, we made the following revisions.


-

This paper describes the application of advanced image processing techniques,

such as

SIDE and SVM, to the detection of solar filaments in H
-
alpha images.

While the methods are novel and the results impressive, the paper is poorly

written and poorly organized. In particular, the use of English is not of a

sufficient level for publication i
n Solar Physics. The authors should therefore

pay more attention to their English before submitting a revised version.


1. Our answer: We reorganize our paper as
follows
.

Section 1: Introduction.

Section 2: Observation.

Section 3: Filament Detection.

3.1
Stabilized Inverse Diffusion Equation.

3.2
Edge Detection Based on Adaptive Thresholding

3.3
Support Vector Machine

3.4
Filament Morphology

3.5
Detection of Filament Disappearance

Section 4: Results.

Section 5: Summary.


To make our methods clear, we add m
ore description for

filament segmentation,

sunspot
detection, filament morphology, and
detection of
filament disappearance in
subsection

3.
2,
3.3,

3.4, and 3.5 respectively. For sunspot detection, we introduce nine features as input to the
SVM classifier.
We also present the results of
segmentation,
sunspot detection, filament
morphology, and filament disappearance in section 4.


Figures are reorganized

and new Figures are added
. Figure 2 in the previous version is
removed according to your comments. Figur
e 3 in the previous version is removed since
readers can find description about the
discontinuous right
-
hand side

of the SIDE from Pollak
et al. (2000). Flow
-
charts of filament detection and filament disappearance detection
are

shown in Figure 2 &
4

respec
tively. Figure 4 & 6 in the previous version are removed.

New
Figure 3 shows the structuring elements of thinning and pruning.

We remove Figure 7 & 8 in
the previous version and put the result of
enhancement

in

Figure
5
. Figure 9 in the previous
version be
comes Figure 6. Figure
7

shows

the result of filament morphology. Figure
8

demonstrates the method of filament disappearance detection.

Final figures:

Figure 1: Filament image.

Figure 2: Flow chart of the filament detection.

Figure 3: Structuring element
s for thinning and pruning.

Figure 4: Flow chart of the detection of filament disappearance.

Figure 5: Enhancement result by the SIDE.

Figure 6: Result of preprocessing, segmentation and sunspot removal.

Figure 7: Filament morphology.

Figure 8: Filament di
sappearance.


English writing has been revised.


-

Section 2 is also too long, and should be shortened significantly. Much

of this section is already published in Pollak et al. (2000) and

Pollak's PhD thesis. Rather than repeating the contents of these

pub
lications, a reference or two would suffice. Shortened versions of

Sections 2.2 and 2.3 would suffice for the purposes of this paper.

This would also make the methods clearer to the Solar Physics

readership. Figure 2 should also be removed as a result.


2.

Our answer: We
rewrite

image enhancement part according to your comments and remove
Figure 2.


-

Section 3.1 on previous image segmentation results should be moved to the

Introduction or Summary sections of the paper, as Section 3 should only

contain meth
ods.


3. Our answer: We move the previous methods to Section 1 introduction.


-

I would also suggest including a figure or flow
-
chart which describes

graphically at which point each image processing task was performed.

This could then be referred to in Sec
tions 2 and 3, making the sequence on

Image

p
rocessing steps used clearer.


4. Our answer: Thank you for your comments. We make two flow
-
charts Figure 2 & 4 to
express our filament detection procedure. Furthermore, we
reorganize

our paper based on the
proc
edure of the flow charts.


-

The title of Section 4 is not consistent with Section titles regularly used

by

Solar Physics. I
would suggest using the title "Results".

"Experimental

Results"

could be misleading to Solar
Physics readers.


5. Our answer: We ma
ke the revision as your comments.


-

Sub
-
section 4.4 entitled Filament Disappearance Report does not contain any

true, verifiable
results in this manuscript. If the authors wish to include

this

discussion, it should be included
in a Discussion or Summary s
ection.


6. Our answer: We develop more progress in filament disappearance. We create the new
subsection 3.5 (Filament Disappearance Detection) to describe the methods for detection of
disappeared filaments and make a flow
-
chart to describe the filament di
sappearance in Figure
4. In subsection 4, we demonstrate the result of the disappearance in Figure 7.


-

All references should be formatted in the Solar Physics style as specified in

the Author Information details on the Solar Physics website.


7. Our answ
er: We are very sorry about our ignorance. We make the revision according to
your comments.


Minor Comments:

-

Addresses should include the country.


8. Our answer: We add country: U.S.A.


-

Abstract:

The use of English in the abstract makes it difficult t
o read and

hence appreciate the
s
ignificance of the work. I would suggest that the

abstract

and indeed the entire paper is
c
orrected for simple English language errors.


9. Our answer: English writing has been revised.


Define SVM as "Support Vector Machin
es (SVM)"


10. Our answer: We make this revision as you suggested.


-

Section 1 (Introduction):


The sentence ending with "... such as structures being supported by

Magnetic

fields" does not make sense. Please re
-
phrase.


11. Our answer: We remove this mea
ningless sentence.


The sentence ending with "... has been proven to be robust" requires

a
reference.


12. Our answer: This sentence has misled readers. Our methods are robust to segment
filaments comparing to the previous methods by Gao et al. and Shih et

al. We test our
methods on many images to show that we can obtain robust results
.


Define SVM in the second paragraph.


13. Our answer:
SVM is defined as Support Vector Machine in the introduction.
In section
3.3 (Support Vector Machines), we add a paragr
aph to define SVM briefly
;

in our experiment,
we use Linear Support Vector Machine. SVM has been introduced by our previous paper (
Qu
et al.
, 2003). R
eaders can
also
refer to Guyon and Stork (2002)

and Vapnik (1998)

for further

information about SVM,


-

Su
b
-
section 2.1 (Linear Multiscale Filtering)

I presume Pollar et al. (2000) should be Pollak et al. (2002).

This error is repeated throughout the paper.


14. Our answer: Thank you for your comments. We correct
this

error. The paper should be
Pollak et al. (
2000) in
IEEE Transactions on Image Processing
.


-

Sub
-
section 3.3 (Adaptive Thresholding Based on Edge Information)


The following statement does not make sense "Since filaments are

usually colder than gaps in the solar chromosphere background

granularity

and therefore appear dark ..." Could the authors please

rephrase? The term "granularity" should probably not be used

with regard to the solar chromosphere. Granules are specific features

visible in the solar photosphere, but not in the chromosphere. The u
se

of the word "granularity" could be misleading to solar physicists.


15. Our answer: We remove this sentence since it is wrong and meaningless.


-

Section 4 (Experimental Results)

Define BBSO in the first sentence.


16. Our answer: We correct it accordin
g to your comments.



Why was a force function of the form given in Equation (12) chosen?

Is there some property
of this equation that makes it highly suitable

for H
-
alpha images or is it simply used because
it
h
as analyzed by

Pollak et al.?


17. Our answe
r: We compared our force function and Pollak et al.
’
s. We found it does not
make any improvement. Therefore, we still use the force function introduced by Pollak et al.
(2000).
We revise this paragraph

in section 3.1

and make this part clear.


Also L is gi
ven in Equation (12), but is not defined here or anywhere

else in the paper.


18. Our answer: We define it as follows.

In the section of result, we define the maximum

neighbor difference L to be 100.


sign(v) is not used regularly in solar physics publica
tions. Please

define "sign".


19. Our answer: We define it according to your comments

in section 3.1
.



-

Sub
-
section 4.3 (Time Complexity)


The title of this section is misleading as this section does not deal with

complexity. Rather, the section describe
s how the programs/algorithms performed

on the authors computer system. This section should be renamed to better

describe the subsection comments.


20. Our answer: We rename it as
“
Computation
al

Time
”
.


The phrase "... CPU time of 1.7 GHz ..." is incorrect
. "1.7

GHz" is

the CPU clock speed.


21. Our answer: We correct it as:
“A computer with an Intel Celeron Processor (2.40GHz)
was used for the filament detection and the development language is IDL.”



"400 x 400 solar images" should include units, i.e., i
s this

“
400x 400" or 400 pixels x 400
pixels?


22. Our answer: We revise it as
400x400

pixels.


-

Sub
-
section 4.4 (Filament Disappearance Report)

"...called: drot_map in the SolarSoftware of the IDL(SSWIDL)" would

be

better written " ... called drot_map in

the SolarSoftWare tree (SSW;

Freeland

and Handy,

1998) ..."


23. Our answer: Thank you for your help. We revise the sentence according to your
comments.