1: \documentclass[12pt,preprint,english]{aastex}

2: \setcounter{tocdepth}{3}

3: \usepackage{graphicx}

4: \usepackage{amssymb}

5: \usepackage{times}

6: 

7: \makeatletter

8: 

9: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% User specified LaTeX commands.

10: \slugcomment{}

11: \shorttitle{Coronal Structure \& Stray Light}

12: \shortauthors{}

13: 

14: \usepackage{babel}

15: \makeatother

16: 

17: \begin{document}

18: 

19: \title{Solar Coronal Structures and Stray Light in \emph{TRACE}}

20: 

21: 

22: \author{C.E. DeForest({*}), P.C.H. Martens(+), and M.J. Wills-Davey(+)}

23: 

24: 

25: \affil{({*})Southwest Research Insitute, 1050 Walnut Street Suite 300, Boulder,

26: CO 80302 USA, deforest@boulder.swri.edu}

27: 

28: 

29: \affil{(+)Smithsonian Astrophysical Observatory}

30: 

31: 

32: \affil{}

33: 

34: \emph{\hfill\large{(In press, Astrophysical Journal, fall 2008)}\hfill}

35: 

36: \begin{abstract}

37: Using the 2004 Venus transit of the Sun to constrain a semi-empirical

38: point-spread function for the \emph{TRACE} EUV solar telescope, we

39: have measured the effect of stray light in that telescope. We find

40: that 43\% of 171\AA\ EUV light that enters \emph{TRACE }is scattered,

41: either through diffraction off the entrance filter grid or through

42: other nonspecular effects. We carry this result forward, via known-PSF

43: deconvolution of \emph{TRACE }images, to identify its effect on analysis

44: of \emph{TRACE }data. Known-PSF deconvolution by this derived PSF

45: greatly reduces the effect of visible haze in the \emph{TRACE }171

46: \AA\ images, enhances bright features, and reveals that the smooth background

47: component of the corona is considerably less bright (and hence more

48: rarefied) than might otherwise be supposed. Deconvolution reveals

49: that some prior conclusions about the Sun appear to have been based

50: on stray light in the images. In particular, the diffuse background

51: {}``quiet corona'' becomes consistent with hydrostatic support of

52: the coronal plasma; feature contrast is greatly increased, possibly

53: affecting derived parameters such as the form of the coronal heating

54: function; and essentially all existing differential emission measure

55: studies of small features appear to be affected by contamination from

56: nearby features. We speculate on further implications of stray light

57: for interpretation of EUV images from \emph{TRACE }and similar instruments,

58: and advocate deconvolution as a standard tool for image analysis with

59: future instruments such as SDO/AIA.

60: \end{abstract}

61: 

62: \keywords{Instrumentation: miscellaneous, Sun: corona, Sun: UV radiation}

63: 

64: 

65: \section{Introduction}

66: 

67: All telescopes, including \emph{TRACE } \citep{Handy1998}, scatter

68: light. The principal scattering mechanisms in a space-based telescope

69: include diffraction through the aperture and any obscuration in the

70: beam of the telescope, irregularities or dust on the mirrors themselves,

71: and reflection or scattering in the detector at the focal plane. All

72: these effects contribute to forming broad, shallow wings on the point-spread

73: function (PSF) of the instrument, which describes the image produced

74: by that instrument when viewing an ideal point source of light. These

75: wings are typically 3-5 orders of magnitude fainter than the core

76: of the PSF, but 3-4 orders of magnitude larger, so that a significant

77: fraction of the light incident into the telescope is spread over a

78: large portion of the image. 

79: 

80: For discrete scenes such as starfields, broad PSF wings are not greatly

81: important except that they reduce the net efficiency of the telescope

82: by reducing the apparent brightness of point sources. For continuous

83: or near-continuous, high-contrast scenes such as viewed by solar telescopes,

84: broad scattering wings are quite important: scattering from large

85: distributed bright structures can overwhelm the emission from dark

86: regions in the same image. These effects, while well known, have received

87: little attention from the solar data analysis community when applied

88: to data from normal-incidence EUV telescopes such as \emph{TRACE,

89: }but they are present nonetheless and must be accounted for when interpreting

90: images from \emph{TRACE }(and all other solar telescopes). Not only

91: quantitative analysis, but even some qualitative interpretations of

92: \emph{TRACE }data may be compromised if stray light effects are not

93: accounted for. 

94: 

95: X rays and ultraviolet light are particularly susceptible to scattering

96: and defocus, and efforts have been made to account for point-spread

97: function effects in previous instruments. For example, \cite{Maute1981}

98: attempted blind iterative deconvolution on the X-ray data from \emph{Skylab,}

99: \cite{svestka1983} applied it to the \emph{SMM/}HXIS instrument;

100: and \cite{Martens1995} determined a spatially variable PSF for the

101: \emph{Yohkoh}/SXT instrument. These studies have largely focused on

102: iterative methods to identify and remove blurring effects caused by

103: a broad PSF core in the subject instruments, though stray light has

104: also been an object of study. Stray light deconvolution was commonly

105: used on SXT data in the later years of that mission (e.g. \citet{Foley1997,Gburek2002,Schrijver2004}). 

106: 

107: The \emph{TRACE} EUV PSF has been studied by several groups. \cite{Lin2001}

108: used compact, bright flares to study diffraction patterns on the \emph{TRACE}

109: focal plane and concluded that diffraction from the aluminum filter

110: grids used in \emph{TRACE} scatters 19\% of incident EUV photons into

111: a highly structured, broad diffraction pattern; they speculated that

112: the scattering may be affecting imaging performance. \cite{gburek2006}

113: used that scattering pattern both to derive both a best-fit PSF core

114: for \emph{TRACE} and also to determine a portion of the emitted EUV

115: spectrum from particular flare events. 

116: 

117: In this report, we consider primarily the diffuse scattering wings

118: of the \emph{TRACE} EUV PSF, and particularly their implications for

119: interpreting coronal images. The wings are not readily measured using

120: a point source such as a flare, because the local intensity of the

121: PSF is quite small far from the core. Outside of diffraction maxima

122: the weak scattered signal is overwhelmed by local emission even for

123: bright events such as flares. Deriving the PSF thus requires analysis

124: of occulted images, using an obstructing body such as the Moon or

125: a planet. We examined \emph{TRACE }data collected near the times of

126: several solar eclipses, but did not find a suitable EUV image set

127: that contained a clear image of the lunar limb. On 2004 Jun 08, Venus

128: passed in front of the Sun, and several 171 \AA\ image sequences were

129: collected as the planet traversed the disk of the Sun and the off-limb

130: corona. We have used those images to derive a semi-empirical scattering

131: PSF for the \emph{TRACE} 171 \AA\ channel, and have tested the PSF for

132: correctness by using it to deconvolve several representative\emph{

133: }images of interesting coronal structures.

134: 

135: In \S\ref{sec:Review-of-Deconvolution} we briefly review deconvolution

136: and how it is performed, in \S\ref{sec:Constraint-of-the} we describe

137: the forward modeling process and present our measured PSF, and in

138: \S\ref{sec:Deconvolution-of-sample} we demonstrate deconvolution

139: of some representative images. Finally, in \S\ref{sec:Discussion}

140: we discuss implications for interpretation of EUV coronal images and

141: recommend deconvolution as a standard reduction pipeline component

142: for future telescopes.

143: 

144: 

145: \section{\label{sec:Review-of-Deconvolution}Review of Deconvolution}

146: 

147: Compensating for the effect of scatter within a telescope requires

148: \emph{deconvolution}: the telescope convolves the scene with the instrument's

149: PSF; the effects of the PSF can then be removed by post-processing.\emph{

150: }Here we briefly review known-PSF deconvolution and how it is performed;

151: the process is much simpler and more robust than {}``blind deconvolution'',

152: which does not require a PSF that is known in advance. 

153: 

154: Telescopes in general respond to a point source of light by generating

155: an image that has finite extent. This image is the PSF of the telescope,

156: and generally varies at most slowly across the image plane; in this

157: treatment, we consider it to be constant with respect to position

158: on the image plane. Images from the telescope are best described as

159: the convolution\emph{ }of the scene being viewed, with the PSF of

160: the telescope. The convolution operation spreads out features by integration

161: (summing) over portions of the source scene, weighted by the PSF.

162: For simplicity, we consider only the post-sampling image plane and

163: use discrete operations such as summing, rather than smooth operations

164: such as integration. 

165: 

166: Convolving an $n_{x}\times n_{y}$ pixel image $I$ with a convolution

167: kernel $K$ (the instrument's PSF) involves taking a weighted sum

168: at each location (i,j) in the source image:

169: 

170: \begin{equation}

171: \left(I\bigotimes K\right)_{i,j}\equiv\sum_{k=-n_{x}}^{n_{y}}\sum_{l=-n_{y}}^{n_{y}}I_{i-k,j-l}K_{k,l}.\label{eq:convolution}\end{equation}

172: By construction, it is clear that convolution is a linear operation,

173: so it can be represented with matrix multiplication of $I$ (treated

174: as a $n_{x}n_{y}$-dimensional column vector) by a $n_{x}n_{y}\times n_{x}n_{y}$

175: matrix $M_{K}$. Undoing the convolution simply requires inverting

176: $M_{K}$.

177: 

178: In general, matrix inversion of large matrices is a hard problem (e.g.

179: \citet{ClaerboutBook}). Fortunately, convolution in real space

180: is equivalent to elementwise multiplication in Fourier space; in other

181: words, the Fourier basis diagonalizes $M_{K}$, so that finding its

182: inverse is trivial. This is the well-known \emph{convolution theorem};

183: a nice treatment and proof may be found in \cite{Bracewell1999}.

184: \begin{equation}

185: I\bigotimes K=\digamma^{-1}\left(\mathcal{I}\cdot\mathbb{\mathcal{K}}\right)\label{eq:conv-theorem}\end{equation}

186:  where $\digamma$ denotes Fourier transformation, the dot product

187: represents elementwise multiplication, and curly vectors $\mathcal{I}$

188: and $\mathcal{K}$ are the Fourier doubles of their italic counterparts

189: $I$ and $K$. The Fourier transform $\mathcal{K}$ of the PSF is

190: the \emph{optical transfer function }of the telescope, and its magnitude

191: $\mathcal{\left|K\right|}$ is the \emph{modulation transfer function}.

192: Inverting the convolution operation, then, just requires multiplying

193: by the reciprocal of the optical transfer function:

194: 

195: \begin{equation}

196: I=\digamma^{-1}(\mathit{\mathcal{I}})=\digamma^{-1}\left(\mathit{\mathit{\mathcal{I\cdot K\cdot}}\mathfrak{R}(\mathit{\mathcal{K}})}\right)\label{eq:inversion}\end{equation}

197: where $\mathfrak{R}$ is the elementwise reciprocal operator. 

198: 

199: Because Fourier transformation is itself a linear operation, the components

200: inside the Fourier transform in Eq. \ref{eq:conv-theorem} can be

201: pulled out, to write:\begin{equation}

202: I=(I\otimes K)\otimes K^{inv}\label{eq:separation}\end{equation}

203: where $I\otimes K$ is a source image, and $K^{inv}$ is the function

204: whose Fourier transform is $\mathfrak{R}(\mathcal{K})$. Eq. \ref{eq:separation}

205: is useful because it shows that the entire deconvolution operation

206: can be represented as a convolution by a single inverse PSF. If the

207: optical transfer function is known, then $K^{inv}$ is trivial to

208: find. It is just:

209: 

210: \begin{equation}

211: K^{inv}=\digamma^{-1}\left(\mathfrak{R}\left(\mathcal{K}\right)\right)\label{eq:finding-kinv}\end{equation}

212: Because of the reciprocal operation, $K^{inv}$ exists only for kernels

213: with no zero Fourier coefficients. 

214: 

215: Even when the PSF $K$ is known, deconvolution is not quite as simple

216: in practice as Eq. \ref{eq:finding-kinv} suggests, because of the

217: presence of noise (which is generally a combination of additive uncorrelated

218: noise, multiplicative uncorrelated noise, and calibration error).

219: A typical image can be represented as a (convolved) true focal-plane

220: image, plus a noise image (which is not wholly independent of the

221: true image): \begin{equation}

222: I=I_{true}\bigotimes K_{}+N\label{eq:noise combination}\end{equation}

223: where $N$ is the noise image. Dividing by $\mathcal{K}_{}$ in the

224: Fourier plane deconvolves the image but also increases the noise term: 

225: 

226: \begin{eqnarray}

227: I_{deconv.} & = & \digamma^{-1}\left(I\cdot\mathfrak{R}\left(\mathcal{K}\right)\right)\label{eq:noise-amplification}\\

228:  & = & I_{true}+\digamma^{-1}\left(\mathcal{N\cdot}\mathfrak{R}\left(\mathcal{K}_{}\right)\right).\label{eq:noise-amp-2}\end{eqnarray}

229: The noise term is left in Fourier space to emphasize that $\mathfrak{R}(\mathcal{K})$

230: is a multiplier on the noise level. Most real telescope PSFs attenuate

231: high spatial frequencies; thus, $\mathfrak{R}(\mathcal{K})$ amplifies

232: those frequencies in the noise by the same factor. The amplified noise

233: term at the far right of Eq. \ref{eq:noise-amp-2} can easily overwhelm

234: $I_{true}$. 

235: 

236: The theory of \emph{Wiener filters }develops the optimal balance between

237: signal amplification and noise suppression for additive noise for

238: linear filters such as deconvolution (e.g. \citet{Press1989}).\emph{

239: }Rather than develop the ideal Wiener filter for each image, it is

240: convenient to prevent excessive noise amplification with a \emph{normalized

241: reciprocal} that rolls over after a certain level of amplification.

242: We used a simple approximation $\mathfrak{R}_{\alpha,\epsilon}$:\begin{equation}

243: \mathfrak{R_{\alpha,\epsilon}}\left(z\right)=\left(\frac{z*}{\left|z\right|}\right)\frac{\left|z\right|^{\alpha}}{\left|z\right|^{\alpha+1}+\epsilon^{\alpha+1}},\label{eq:normalized-recip}\end{equation}

244:  where $z$ is complex but $\alpha$ and $\epsilon$ are real. $\mathfrak{R}_{\alpha,\epsilon}$

245: converges to $z^{-1}$ for large values of $z$ and to $z*\left|z\right|^{\alpha-1}\epsilon^{-\alpha-1}$

246: for small values of $\left|z\right|$ (when compared with $\epsilon$),

247: and rises to a maximum value close to $\alpha\epsilon^{-1}$. Figure

248: \ref{fig:Inversion} demonstrates construction of an inverse kernel

249: using $\mathfrak{R}_{0.5,0.05}$ for the reciprocal. 

250: 

251: %

252: \begin{figure}[!tbh]

253: \includegraphics[width=6in]{f1.ps}

254: 

255: \caption{\emph{\label{fig:Inversion}Inversion of a sample circularly symmetric kernel

256: using the Fourier transform and regularized reciprocation. (A): original

257: sample PSF, $K$; (B): true (dashed) and regularized (solid) reciprocal

258: of the Fourier components of $K$; (C): resulting inverse kernel $K^{inv}$;

259: (D): convolution of $K\bigotimes K^{inv}$ would be a delta function

260: in the ideal case; it is much improved over the original in (A), which

261: is overplotted as a dashed line.}}

262: 

263: \end{figure}

264: 

265: \section{\label{sec:Constraint-of-the}Constraint of the \emph{TRACE} PSF }

266: 

267: We generated a PSF using data from the Venus transit of 2004 June

268: 08, assuming that Venus emits no EUV light. The images were prepared,

269: cleaned, and aligned, and a forward model of the scattering was made

270: by convolving the EUV solar images with a PSF to determine the effect

271: of the emission on the center of the Venus image. We generated a forward

272: model PSF that had a narrow core (because we were not interested in

273: sharpening the images, only in reducing stray light) and included

274: \emph{a priori} the known 171 \AA\ diffraction pattern first described

275: by \cite{Lin2001}. Added to the core and the diffraction pattern

276: was a broad truncated Lorentzian described by three parameters: the

277: height, the width parameter, and the width of a Gaussian envelope

278: that was used to truncate the Lorentzian. We convolve the parameterized

279: PSF with the solar EUV images (with the portion inside the disk of

280: Venus masked to black) and calculated a model intensity at each of

281: twelve test loci within the disk of Venus: the center of each of six

282: images of the planet, and two offset loci in each of three on-disk

283: images. We compared these model intensities to the original image

284: brightnesses at the test loci, and adjusted the parameters to find

285: the best fit for all twelve loci.

286: 

287: 

288: \subsection{Image preparation}

289: 

290: Two sets of \emph{TRACE} EUV images are present from the Venus occultation,

291: both in the 171 \AA\ passband. During the transit itself many images

292: were taken with no binning and either 16 s, 30 s, or 90 s exposure

293: (in the hour 09:00 - 10:00 UT); and shortly after the transit a series

294: of images with 16 second exposure and 2x2 binning were collected off-limb

295: (in the hour 11:30 - 12:30 UT). On-disk, we used the 90 s exposures

296: to minimize background noise; off-limb, we median-filtered blocks

297: of 5 images along the time axis, to reduce background noise in those

298: less-well-exposed frames. Figure \ref{fig:TRACE raw images} shows

299: a sample on-disk and off-limb image. We downloaded Level 1 data directly

300: from the \emph{TRACE} web site. We further corrected the zero point

301: by subtracting the average of the 30x30 pixel region in the lower-left

302: corner of each \emph{TRACE} image (in the filter vignetted area),

303: then scaled the images to 1'' per pixel (thus binning 2x2 the on-disk

304: images) and divided out the exposure time and binned pixel size to

305: arrive at calibrated images in \emph{TRACE} digitizer numbers (DN)

306: arcsec$^{-2}$ s$^{-1}$, so that the images were directly comparable.

307: We despiked each image using a simple unsharp-mask + threshold algorithm

308: (the {}``spikejones.pdl'' routine in the PDL portion of the \emph{Solarsoft}

309: (\citet{FreelandHandy1998}) software distribution), and replaced

310: each spike value with the median of valid values in its neighborhood

311: in the same image. We made a 5-image pixelwise median of the off-limb

312: images, then spatially shifted each image to center Venus in the frame

313: (Figure \ref{fig:Venus-regularize}). 

314: 

315: Because \emph{TRACE} has a limited field of view, but the primary

316: and secondary mirrors are exposed to the entire solar disk, it was

317: necessary to extend the field of view to a good fraction of the solar

318: disk to model the extended PSF and reduce the possibility that edge

319: effects would affect the result. We used the closest-in-time full-disk

320: 171\AA\ image from \emph{SOHO}/EIT (\citet{Delaboudiniere1995}), collected

321: at approximately 19:00 UT on the same day, to fill in missing values

322: outside the \emph{TRACE} field of view. Although the EIT image was

323: collected some 7-9 hours after the Venus occultation data, the portions

324: of the solar image that are affected are far from test loci in the

325: disk of Venus, and therefore only large spatial scales are important;

326: brightness on these scales varies on timescales of hours to days.

327: The EIT image was prepared using the instrument-supplied eit\_prep

328: software, scaled to 1'' per pixel, derotated to the \emph{TRACE} time,

329: and multiplied so that a 100x100 pixel sum (chosen to be far from

330: the \emph{TRACE} vignetted regions) was equal between the EIT and

331: \emph{TRACE} image. Then the dark (vignetted) portions of the \emph{TRACE}

332: image were replaced with the corresponding portion of the scaled,

333: corrected EIT image. Figure \ref{fig:Venus-regularize} shows all

334: of the resulting cleaned, combined images and the loci within them

335: that were used to constrain the fit.

336: 

337: %

338: \begin{figure}[!htb]

339: \includegraphics[width=6in]{f2.ps}

340: 

341: \caption{\emph{\emph{TRACE} EUV images of the 2004 Venus transit (left) and its immediate

342: aftermath (right), showing the restricted size of the on-disk field

343: of view. The images were extended by combining them with the closest

344: EIT image at the same wavelength. \label{fig:TRACE raw images}}}

345: 

346: \end{figure}

347: 

348: 

349: 

350: %

351: \begin{figure}[!tb]

352: \center{\includegraphics[height=6.5in]{f3.ps}}

353: 

354: \caption{\emph{\label{fig:Venus-regularize}Regularized images from the 2004 June

355: 08 Venus transit used for fitting the \emph{TRACE} scattering PSF.

356: The EIT image from 19:00 has been used to fill in the missing regions

357: outside the original field of view. The field of view is 800'' across;

358: the best-fit PSF has a scattering full-width of \textasciitilde{}300''.

359: The sample regions are marked: each image is sampled in a 12''x12''

360: square at the center of the (58'' diameter) disk of Venus, and disk

361: images are also sampled in a 12''x6'' rectangle offset up 18'' from

362: the center of the disk, and 6''x12'' rectangle offset left 18'' from

363: the center of the disk.}}

364: 

365: 

366: 

367: \end{figure}

368: 

369: 

370: 

371: \subsection{Forward Modeling of the PSF}

372: 

373: We forward modeled the scattering portion of the \emph{TRACE} PSF

374: as the sum of the measured diffraction pattern \citep{Lin2001,gburek2006}

375: and a circularly symmetric scattering profile produced by revolving

376: a radial function about the origin. The revolved function was a sum

377: of a narrow core, a {}``shoulder'' Gaussian, and a truncated Lorentzian

378: intended to represent the scattering wings. The central core width

379: was chosen to have a value much less than the Gburek et al. (2006)

380: width, because the intent is to remove scattering wings rather than

381: to sharpen the core of the PSF by deconvolution. The diffraction pattern

382: was convolved with the central core to avoid pixelization artifacts

383: due to the delta functions in it.

384: 

385: The analytic formula is:\begin{equation}

386: K_{\alpha,w_{},\sigma}(r,\theta)=\gamma^{-1}\left(\left(e^{-4ln(2)r^{2}}/1.27\right)\otimes D(r,\theta)+\frac{\alpha}{r^{2}/w^{2}+1}e^{-4ln(2)r^{2}/\sigma_{t}^{2}}+\beta e^{-4ln(2)r^{2}/\sigma_{s}^{2}}\right),\label{eq:PSF}\end{equation}

387: where $r$ is distance in the image plane, measured in arcseconds;

388: the 1.27 normalizes the integral under the first Gaussian to unity;

389: $\otimes$ represents convolution; $D(r,\theta)$ is the diffraction

390: pattern described below, including a central core; $\alpha$ is relative

391: strength of the Lorentzian wings; $w$ is the width of the Lorentzian;

392: $\sigma_{t}$ is the FWHM of the truncating Gaussian (in arcsec);

393: $\beta$ is the strength of a Gaussian shoulder to the curve; $\sigma_{s}$

394: is the width of the shoulder; and $\gamma$ is a factor to normalize

395: the integral under the 2-D convolution kernel to unity. 

396: 

397: The parameters were found by an iterative fit method: for each guess

398: set of parameters, the kernel was calculated on an 800''x800'' grid

399: at 0.5'' resolution, multiplied by each of the six composite images,

400: and summed to find the expected scattered intensity at the center

401: of each Venus image. Note that the \emph{TRACE} field of view is only

402: about 500'' across -- the larger FOV (available with the EIT overlay)

403: was used so that the \emph{TRACE} field of view, itself, wouldn't

404: constrain the fits. 

405: 

406: In addition, we used an offset kernel to calculate the intensity at

407: an off-center locus in each of the three on- disk images, to constrain

408: the shoulders of the curve a few arcseconds from the core. We did

409: not use the off-center brightness in the off-limb Venus images, because

410: pixels above the limb of the Sun probably contain proportionally more

411: scattered light than do pixels on the solar disk. The geometry of

412: each sample point is shown in Figure \ref{fig:Venus-regularize}.

413: We compared the intensities to the forward scattering model, and adjusted

414: the parameters initially {}``by eye'' to find a reasonable match

415: with the twelve data points. Finally, we optimized the fit with an

416: amoeba algorithm (e.g. \citet{Pressbook}), holding $w_{core}$

417: at the conservative 0.5 $arcsec$ full-width at half-maximum and penalizing

418: errors in the overcompensation direction (taking the image value below

419: zero) a factor of 100x worse than errors in the undercompensation

420: direction. The resulting parameters are given by:

421: 

422: \begin{equation}

423: \begin{array}{ccc}

424: \alpha & = & 2.06\times10^{-5}\\

425: w & = & 57.7\, arcsec\\

426: \sigma_{t} & = & 68.4\, arcsec\\

427: \beta & = & 6\times10^{-7}\\

428: \sigma_{s} & = & 15\, arcsec\end{array}\label{eq:best-fit-params}\end{equation}

429: 

430: 

431: The fit is within $0.021\, ct\, arcsec^{-2}\, s^{-1}$ of the measured

432: data value at each sample point and 0.019 $ct\, arcsec^{-2}\, s^{-1}$

433: RMS across the 12 data points, compared to absolute brightnesses of

434: 0.5-1 $ct\, arcsec^{-2}\, s^{-1}$ in the interior of the disk of

435: Venus. The fitted PSF and its inverse are plotted in Figure \ref{fig:PSF}.

436: There is no significant contribution to the total energy outside of

437: a 100'' radius. The maximum intensity in the PSF core is 0.5; hence,

438: the modeled isotropic scattering function is down by 5 orders of magnitude

439: from the center of the PSF. Nevertheless, its large cover compared

440: to the core of the PSF yields a significant amount of scattering.

441: 

442: %

443: \begin{figure}[!htb]

444: \center{\includegraphics[bb=0bp 0bp 240bp 226bp,clip,width=4in]{f4.ps}}

445: 

446: \caption{\emph{\label{fig:PSF}The best-fit point-spread function (and its inverse)

447: for the 171\AA\ channel of \emph{TRACE}. The intensity is plotted on

448: a log scale and is normalized to an integral of unity. }}

449: 

450: \end{figure}

451: 

452: 

453: The encircled energy is plotted versus distance in Figure \ref{fig:Encircled-energy}.

454: The encircled energy curve can be counter-intuitive at first: while

455: the scattering wings start and remain small (at under $10^{-3}$ of

456: the intensity at the core of the PSF), at each successive radius more

457: area is available to contribute to the total integrated energy. Hence,

458: in a nearly uniform scene most of the stray light at a given point

459: in the image plane arises from features 20-50 $arcsec$ (40-100 \emph{TRACE

460: }pixels) away. Approximately 43\% of the energy in the derived PSF

461: exists more than 2'' from the center of the core.

462: 

463: \begin{figure}[!htb]

464: \center{\includegraphics[width=3in]{f5.ps}}

465: 

466: \caption{\emph{\label{fig:Encircled-energy}Encircled energy vs. distance for the

467: model PSF calculated with Equations \ref{eq:PSF}-\ref{eq:best-fit-params}.

468: About 57\% of the PSF's total energy is contained in the central core,

469: and 43\% is contained in broad scattering wings that extend about

470: 100'' (200 \emph{TRACE} pixels) in all directions. The {}``wobbles''

471: in the curve are due to successive diffraction maxima.}}

472: \end{figure}

473: 

474: 

475: 

476: \section{\label{sec:Deconvolution-of-sample}Deconvolution of sample \emph{TRACE}

477: images}

478: 

479: Scattering of 43\% has a significant impact on images collected with

480: \emph{TRACE.} Here we present results of deconvolution, using the

481: measured PSF from \S\ref{sec:Constraint-of-the}. A broad variety

482: of 171\AA\ images were deconvolved and tested for correctness. In no

483: case did any of the pedestal-subtracted, deconvolved images have significant

484: negative-flux regions, an indication that the fitted PSF is either

485: correct or conservative compared to the real PSF of the instrument.

486: Isolated pixels may be carried below zero, due to JPEG artifacts or

487: photon counting noise, but smoothing the image with a $5\times5$

488: pixel boxcar kernel eliminates the negative regions. In general, bright

489: features get marginally brighter, and dark regions get much darker,

490: after deconvolution. Figure \ref{fig:deconvolution-demo} shows the

491: results of deconvolution of a limb scene, a near-disk-center scene,

492: and several dark prominences near the limb. All of the images have

493: been cropped to the middle $500\times500$ pixels of the \emph{TRACE

494: }detector after deconvolution, to avoid edge effects.

495: 

496: The images were despiked using a spatial spike finder

497: ({}``spikejones.pdl'' in the \emph{Solarsoft }distribution; \citet{FreelandHandy1998})\emph{

498: }In each case, the median value of the lower, left 15x15 pixel region

499: from each despiked \emph{TRACE} image was used as a zero-point reference:

500: because \emph{TRACE }is vignetted by the thin-foil filter ring on

501: board, the lower-left corner serves as a reasonable dark reference

502: value. No additional background subtraction was performed.

503: 

504: In general, contrast is greatly enhanced throughout the images. For

505: example, the lane in the disk-center active region (center row of

506: Figure \ref{fig:deconvolution-demo}) is shown to be about a factor

507: of 3 darker than might be expected from na\"ive analysis of the image

508: (without deconvolution), and small features embedded in bright regions

509: (such as the fan of threads on the right hand side of that image)

510: can be as much as doubled in contrast relative to their local background.

511: The prominences in the bottom row of Figure \ref{fig:deconvolution-demo}

512: demonstrate the effectiveness of deconvolution at removing nearby

513: coronal brightness: the prominences (which protrude about one density

514: scale height at 1MK, hence two intensity scale heights) are shown

515: to be quite dark, presumably because they protrude above most of the

516: quiet coronal emissions. The leftmost prominence is directly behind

517: a loop base and is therefore not darkened nearly as much as the others;

518: this forms a good check that the deconvolution is not simply darkening

519: features arbitrarily. The second-from-left prominence is seen between

520: two bright loop structures, and is hence quite dark despite the apparent

521: bright foreground.

522: 

523: %

524: \begin{figure}[!tb]

525: \includegraphics[width=6in]{f6.ps}

526: 

527: \caption{\emph{\label{fig:deconvolution-demo}Deconvolution of sample \emph{TRACE}

528: 171 images greatly increases contrast of dark features, eliminates

529: `haze'. Top: limb scene. Center: disk AR. Bottom: Prominences near

530: the limb. In each row, left is a level-1 processed \emph{TRACE }image;

531: middle is the same image, deconvolved; and right is a plot of brightness

532: along the indicated cut. Compact bright features (such as the small

533: active region at the left hand side of the top panel) are increased

534: by nearly 50\% in brightness; dark features in the midst of bright

535: regions (such as the dark lane near the center of the central image,

536: or the dark prominences at bottom) are darkened by a factor of about

537: 3. Note that the prominence at far left of the lower panel is behind

538: a large bright loop structure, while the others in that image are

539: not.}}

540: 

541: \end{figure}

542: 

543: 

544: 

545: \section{\label{sec:Discussion}Discussion}

546: 

547: The result that \emph{TRACE} images contain a significant amount of

548: scattered light is not, in itself, new. Most telescope PSFs include

549: scattering wings, and \emph{TRACE }is no exception. The filter grid

550: in the front of the telescope is known to scatter $\sim20\%$ of incident

551: light into a structured diffraction pattern with myriad local maxima

552: (\cite{Lin2001,gburek2006}). The present analysis is new in three

553: important respects: it is (to our knowledge) the first analysis of

554: scattered light using \emph{TRACE }occultation data to derive a PSF;

555: much more scattered light is found than can be accounted for merely

556: by diffraction; and the process is taken to its natural conclusion

557: of deconvolving the original \emph{TRACE }images to show the effect

558: of the stray light on scientific interpretation of the images. 

559: 

560: The \emph{TRACE }Venus data are not detailed enough to constrain a

561: highly structured PSF model, but the simple empirical fit described

562: here is sufficient to improve existing images via deconvolution, and

563: passes the most basic of deconvolution tests, suggesting that it is

564: not overcompensating for the scattering wings. Deconvolution with

565: our scattering PSF has a similar effect to background subtraction

566: on the interpretation of small features: for features small compared

567: to the scattering wings, the effect of the surrounding bright features

568: is approximately constant, so subtraction of a modeled or fixed background

569: yields similar effects in particular local areas of a given \emph{TRACE

570: }image. The principal advantages of deconvolution are that an approximation

571: of absolute brightness is reproduced, rather than the offset relative

572: brightness that may be extracted from simple background-subtracted

573: images; and that moderate-scale features are treated correctly (they

574: are not treated correctly by simple background or pedestal subtraction). 

575: 

576: Deconvolution not only darkens the faintest portions of the image,

577: it both increases the relative contrast of small bright features embedded

578: in a bright background and also affects photometric estimates of the

579: relative density of any small bright feature seen with \emph{TRACE}.

580: This describes several features of interest in the \emph{TRACE} data,

581: including active-region threads that are a subject of current debate

582: \citep{WatkoKlimchuk2000,WarrenWinebarger2003,Fuentes2006,DeForest2007}. 

583: 

584: The greater coronal contrast we find in deconvolved \emph{TRACE} images

585: gives indirect support to the idea that the corona is close to hydrostatic

586: equilibrium despite the observed tallness of bright features such

587: as active region loops. The coronal density scale height is about

588: 50 Mm at 1 MK, so the emissivity scale height of Fe IX \& Fe X emission

589: line features (close to 1MK ionization temperature) is about 25 Mm

590: (0.035 $R_{S}$) assuming local thermal equilibrium. Thus the EUV-visible

591: corona might be expected to form a thin layer near the photosphere

592: with no significant emission arising at altitudes higher than about

593: 0.07 $R_{S}$. Essentially all \emph{TRACE }EUV images show significant

594: background brightness high in the corona; the brightness is visible

595: above the detector {}``pedestal'', because there is contrast between

596: dark but {}``live'' pixels that are part of the image, and corner

597: pixels that are vignetted by the round filter mount at the back of

598: the instrument. The current measured instrument PSF suggests that

599: most or all of this background brightness is due to scattering within

600: the telescope, because dark features (such as the prominences in the

601: bottom row of Figure \ref{fig:deconvolution-demo}) are reduced nearly

602: to zero brightness when deconvolved. This is the general behavior

603: to be expected from a thin hydrostatic atmosphere at a particular

604: temperature: tall features that are more than 1-2 scale heights tall

605: should have little or no emission above them.

606: 

607: Active region loops appear to have a large scale height compared to

608: that expected for 1-2 MK plasma \citep{SchrijverMcMullen2000,AschwandenNitta2000,WinebargerWarrenMariska2003,Fuentes2006,DeForest2007}.

609: Three explanations that have been advanced are resonant scattering

610: of EUV (which varies as $n_{e}$ rather than $n_{e}^{2}$), support

611: by non-hydrostatic momentum transport mechanisms such as siphon flows

612: or wave motion, or geometric considerations that attenuate brightness

613: at the bases of the loop. Our result that the quiet corona appears

614: to be consistent with the expected hydrostatic scale height seems

615: to eliminate resonant scattering as a mechanism for tallness, because

616: it would imply a stronger haze in the foreground at high solar altitudes.

617: Further, it seems to limit the functional form of anomalous support

618: mechanisms that could lengthen active region loops' scale height,

619: because such mechanisms must act preferentially against active region

620: loops and not quiet sun loops, to be consistent with the morphology

621: of the deconvolved images. This can further be construed as circumstantial

622: evidence for a geometric, rather than intrinsic, explanation for active

623: region loops' long apparent length \citep{DeForest2007}. 

624: 

625: \cite{Moore2008} have recently used image-processing techniques

626: to separate the hazy and sharp components of active region loops viewed

627: with \emph{TRACE. }Such analyses rely on the sharp component of the

628: corona as an indicator of stray light. With 43\% scattering of the

629: total light incident on the telescope, 57\% is left to be focused;

630: hence, we expect that the hazy portion of such separated image pairs

631: derived from \emph{TRACE }171\AA\  data should contain about 75\%

632: as much total brightness as the sharp portion does, on the basis of

633: stray light alone. 

634: 

635: In addition to morphological differences, corrections to the relative

636: brightness of features such as active region threads and voids affect

637: parameters such as the derived Alfv\'en speed, because of the $n_{e}^{2}$

638: dependence of EUV emission. Structures with spiky density profiles

639: emit more EUV per electron than do smooth structures, and the inferred

640: Alfv\'en speed depends both on the magnetic field and the derived electron

641: density. Onset of some \emph{TRACE-}observed EIT waves appears to

642: require high Alfv\'en speeds of up to $3\, Mm\, s^{-1}$ \citep{willsdavey2007};

643: this high speed is difficult to explain in the presence of a diffuse

644: background corona around the source region of the EIT wave. If in

645: fact active regions contain nearly evacuated regions (as in the center

646: panel of Figure \ref{fig:deconvolution-demo}), then the variation

647: in Alfv\'en speed is greatly increased and the region-wide average Alfv\'en

648: speed may be significantly higher than would otherwise be inferred.

649: 

650: As a final example of the impact of stray light in EUV images\emph{,

651: }coronal heating properties have been derived \citep{Schrijver2004}

652: by examining the contrast between coronal holes and bright structures,

653: and may be affected by scattering in \emph{SOHO}/EIT and/or \emph{Yohkoh}/SXT.

654: Specifically, if coronal holes are significantly darker, and bright

655: structures are significantly sharper and brigher, than is apparent

656: in raw EUV and X-ray images, then the coronal heating mechanism may

657: not be as distributed as might otherwise be inferred.

658: 

659: The model PSF that we have derived for \emph{TRACE} is somewhat simplified:

660: we have included, \emph{a priori, }detailed structure that is known

661: from earlier studies \citep{Lin2001}, and parameterized

662: an additional scattering term based on the empirical behavior of stray

663: light on rough mirrors. We have not taken into account possible anisotropy

664: or spatial variability of the scattering, attempted to gain physical

665: understanding of the causes of the PSF, or modeled scattering phenomena

666: that do not fit within the paradigm of a simple PSF. Based on measurements

667: of the Venus transit in 2004, we have found that roughly 43\% of incident

668: energy is scattered by \emph{TRACE}, so that approximately half of

669: the scattered energy may be ascribed to the diffraction pattern found

670: by Lin et al. and approximately half to other mechanisms. Deconvolution

671: greatly improves contrast in \emph{TRACE }images, raising concerns

672: about the interpretation of those images.

673: 

674: More generally, deconvolution to increase contrast in images with

675: scattering wings is strongly recommended for observation from present

676: and future EUV and X-ray telescopes. We have shown that deconvolution

677: can greatly affect the contrast of observed features, and discussed

678: how this effect may affect a broad variety of science questions. Further,

679: deconvolution to remove broad scattering wings is in general not as

680: hazardous to the data as is deconvolution to increase sharpness in

681: the core of the telescope PSF. That is because there are high spatial

682: frequencies present in the core of the kernel, even if it is added

683: to a much broader distribution, so that noise is not increased as

684: much from deconvolution of a scattering PSF as from a broad PSF core.

685: 

686: 

687: \acknowledgements{Thanks to the \emph{TRACE} and \emph{SOHO}/EIT teams for making their

688: data available to everyone. We also owe thanks to L. Golub, J. Cirtain,

689: K. Schrijver, and H. Throop for illuminating discussions, and to the

690: anonymous referee for several suggestions that improved the work.

691: SOHO is a project of international cooperation between NASA and ESA.

692: This work was funded under NASA's SHP-SR\&T program.}

693: 

694: \begin{thebibliography}{24}

695: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi

696: 

697: \bibitem[{{Aschwanden} \& {Nitta}(2000)}]{AschwandenNitta2000}

698: {Aschwanden}, M.~J. \& {Nitta}, N. 2000, \apjl, 535, L59

699: 

700: \bibitem[{{Bracewell}(1999)}]{Bracewell1999}

701: {Bracewell}, R.~N. 1999, {The Fourier Transform and Its Applications}

702:   (McGraw-Hill)

703: 

704: \bibitem[{{Claerbout}(2004)}]{ClaerboutBook}

705: {Claerbout}, J. 2004, {Image Estimation by Example} (Stanford Exploration

706:   Project. Available online: http://sepwww.stanford.edu/sep/prof/gee)

707: 

708: \bibitem[{{DeForest}(2007)}]{DeForest2007}

709: {DeForest}, C.~E. 2007, \apj, 661, 532

710: 

711: \bibitem[{{Delaboudiniere} {et~al.}(1995){Delaboudiniere}, {Artzner},

712:   {Brunaud}, {Gabriel}, {Hochedez}, {Millier}, {Song}, {Au}, {Dere}, {Howard},

713:   {Kreplin}, {Michels}, {Moses}, {Defise}, {Jamar}, {Rochus}, {Chauvineau},

714:   {Marioge}, {Catura}, {Lemen}, {Shing}, {Stern}, {Gurman}, {Neupert},

715:   {Maucherat}, {Clette}, {Cugnon}, \& {van Dessel}}]{Delaboudiniere1995}

716: {Delaboudiniere}, J.-P., {Artzner}, G.~E., {Brunaud}, J., {Gabriel}, A.~H.,

717:   {Hochedez}, J.~F., {Millier}, F., {Song}, X.~Y., {Au}, B., {Dere}, K.~P.,

718:   {Howard}, R.~A., {Kreplin}, R., {Michels}, D.~J., {Moses}, J.~D., {Defise},

719:   J.~M., {Jamar}, C., {Rochus}, P., {Chauvineau}, J.~P., {Marioge}, J.~P.,

720:   {Catura}, R.~C., {Lemen}, J.~R., {Shing}, L., {Stern}, R.~A., {Gurman},

721:   J.~B., {Neupert}, W.~M., {Maucherat}, A., {Clette}, F., {Cugnon}, P., \& {van

722:   Dessel}, E.~L. 1995, \solphys, 162, 291

723: 

724: \bibitem[{{Foley} {et~al.}(1997){Foley}, {Culhane}, \& {Acton}}]{Foley1997}

725: {Foley}, C.~R., {Culhane}, J.~L., \& {Acton}, L.~W. 1997, \apj, 491, 933

726: 

727: \bibitem[{{Freeland} \& {Handy}(1998)}]{FreelandHandy1998}

728: {Freeland}, S.~L. \& {Handy}, B.~N. 1998, \solphys, 182, 497

729: 

730: \bibitem[{{Fuentes} {et~al.}(2006){Fuentes}, {Klimchuk}, \&

731:   {D{\'e}moulin}}]{Fuentes2006}

732: {Fuentes}, M.~C.~L., {Klimchuk}, J.~A., \& {D{\'e}moulin}, P. 2006, \apj, 639,

733:   459

734: 

735: \bibitem[{{Gburek} \& {Sylwester}(2002)}]{Gburek2002}

736: {Gburek}, S. \& {Sylwester}, J. 2002, \solphys, 206, 273

737: 

738: \bibitem[{{Gburek} {et~al.}(2006){Gburek}, {Sylwester}, \&

739:   {Martens}}]{gburek2006}

740: {Gburek}, S., {Sylwester}, J., \& {Martens}, P. 2006, \solphys, 239, 531

741: 

742: \bibitem[{{Handy} {et~al.}(1998){Handy}, {Bruner}, {Tarbell}, {Title},

743:   {Wolfson}, {Laforge}, \& {Oliver}}]{Handy1998}

744: {Handy}, B.~N., {Bruner}, M.~E., {Tarbell}, T.~D., {Title}, A.~M., {Wolfson},

745:   C.~J., {Laforge}, M.~J., \& {Oliver}, J.~J. 1998, \solphys, 183, 29

746: 

747: \bibitem[{{Lin} {et~al.}(2001){Lin}, {Nightingale}, \& {Tarbell}}]{Lin2001}

748: {Lin}, A.~C., {Nightingale}, R.~W., \& {Tarbell}, T.~D. 2001, \solphys, 198,

749:   385

750: 

751: \bibitem[{{Martens} {et~al.}(1995){Martens}, {Acton}, \& {Lemen}}]{Martens1995}

752: {Martens}, P.~C., {Acton}, L.~W., \& {Lemen}, J.~R. 1995, \solphys, 157, 141

753: 

754: \bibitem[{{Maute} \& {Elwert}(1981)}]{Maute1981}

755: {Maute}, K. \& {Elwert}, G. 1981, \solphys, 70, 273

756: 

757: \bibitem[{{Moore} {et~al.}(2008){Moore}, {Cirtain}, \& {Falconer}}]{Moore2008}

758: {Moore}, R.~L., {Cirtain}, J.~W., \& {Falconer}, D.~A. 2008, \emph{Eos Trans.

759:   AGU}, 89(23), Jt. Assem. Suppl, Abstract SP43C

760: 

761: \bibitem[{{Press} {et~al.}(1989){Press}, {Flannery}, {Teukolsky}, \&

762:   {Vetterling}}]{Press1989}

763: {Press}, W.~H., {Flannery}, B.~P., {Teukolsky}, S.~A., \& {Vetterling}, W.~T.

764:   1989, {Numerical Recipes} (Cambridge University Press)

765: 

766: \bibitem[{{Press} {et~al.}(2007){Press}, {Teukolsky}, {Vetterling}, \&

767:   {Flannery}}]{Pressbook}

768: {Press}, W.~H., {Teukolsky}, S.~A., {Vetterling}, W.~T., \& {Flannery}, B.~P.

769:   2007, Numerical Recipes 3rd Edition: The Art of Scientific Computing

770:   (Cambridge University Press)

771: 

772: \bibitem[{{Schrijver} \& {McMullen}(2000)}]{SchrijverMcMullen2000}

773: {Schrijver}, C.~J. \& {McMullen}, R.~A. 2000, \apj, 531, 1121

774: 

775: \bibitem[{{Schrijver} {et~al.}(2004){Schrijver}, {Sandman}, {Aschwanden}, \&

776:   {DeRosa}}]{Schrijver2004}

777: {Schrijver}, C.~J., {Sandman}, A.~W., {Aschwanden}, M.~J., \& {DeRosa}, M.~L.

778:   2004, \apj, 615, 512

779: 

780: \bibitem[{{Svestka} {et~al.}(1983){Svestka}, {Schrijver}, {Somov}, {Dennis},

781:   {Woodgate}, {Fuerst}, {Hirth}, {Klein}, \& {Raoult}}]{svestka1983}

782: {Svestka}, Z., {Schrijver}, J., {Somov}, B., {Dennis}, B.~R., {Woodgate},

783:   B.~E., {Fuerst}, E., {Hirth}, W., {Klein}, L., \& {Raoult}, A. 1983,

784:   \solphys, 85, 313

785: 

786: \bibitem[{{Warren} \& {Winebarger}(2003)}]{WarrenWinebarger2003}

787: {Warren}, H.~P. \& {Winebarger}, A.~R. 2003, \apjl, 596, L113

788: 

789: \bibitem[{{Watko} \& {Klimchuk}(2000)}]{WatkoKlimchuk2000}

790: {Watko}, J.~A. \& {Klimchuk}, J.~A. 2000, \solphys, 193, 77

791: 

792: \bibitem[{{Wills-Davey} {et~al.}(2007){Wills-Davey}, {Sechler}, \&

793:   {McIntosh}}]{willsdavey2007}

794: {Wills-Davey}, M.~J., {Sechler}, M., \& {McIntosh}, S.~W. 2007, AGU Fall

795:   Meeting Abstracts, A222+

796: 

797: \bibitem[{{Winebarger} {et~al.}(2003){Winebarger}, {Warren}, \&

798:   {Mariska}}]{WinebargerWarrenMariska2003}

799: {Winebarger}, A.~R., {Warren}, H.~P., \& {Mariska}, J.~T. 2003, \apj, 587, 439

800: 

801: \end{thebibliography}

802: 

803: 

804: \end{document}

805: