next previous contents

SECTION 3 - RAW-IMAGE SCREENING OPERATIONS

Certain operations performed by IUESIPS may be categorized as screening or preprocessing operations which prepare the data for subsequent processing. The operations of this nature which are currently performed on IUE data are described in this section.

3.1 MICROPHONIC NOISE DETECTION

Since the fall of 1981 at GSFC (28 September 1981 for low dispersion, 10 November 1981 for high dispersion), LWR images have been preprocessed to detect the presence of periodic noise interference ("microphonic noise"). In the SWP and LWP cameras, interference often covers all of the image and its amplitude is generally low (often only 1-3 DN) compared to random background noise (Northover, 1980), making detection difficult. In the LWR, however, the microphonic noise has different characteristics, being localized chiefly in a small number of image lines, well-modeled by an exponentially damped sinusoid, and with a peak amplitude typically in excess of 20 DN (Northover, 1980; Panek and Schiffer, 1981). The LWR microphonic noise is descriptively referred to as a "ping" because of its sudden onset and rapid decay. Unless an extended heater warm-up prior to image read is used as a ping avoidance technique, a given LWR image has about an 85 percent probability of suffering a ping, generally in the lower one-third of the image (Holm and Panek, 1982). The extended heater warm-up ping avoidance technique generally displaces the ping off the image or to the upper part of the image.

Utilizing techniques developed at VILSPA by K. Northover (1980), the microphonics screening done by IUESIPS for all LWR images is based on the characteristics of the image data in the last 32 samples of each image line. This area is outside of the target region, where the pixel values are zero except for noise. With the IUESIPS program MICRO, lines affected by the microphonics are initially identified by applying a threshold to the variance of the last 32 samples; the interference amplitude is then estimated on the basis of the power spectrum of the sampled data, with successive image lines processed in pairs. Image lines with estimated noise amplitudes in excess of the chosen threshold (corresponding to a peak-to-peak noise amplitude of f 10 DN) are flagged by a notation in the image processing history portion of the label (see Section 9.3 ), which is subsequently used by other programs: the spectral registration routine (Section 6.3.2.1 ) uses this information to avoid areas affected by the microphonics, and the spectral extraction routines (Section 7 ) use this information to flag extracted fluxes derived from lines affected by the microphonic interference.

3.2 BRIGHT-SPOT DETECTION

Long IUE exposures characteristically contain "bright spots", i.e., pixels with unusually high DN values which comprise discrete impulse noise often reaching the saturation level. Such bright spots are thought to be caused either by permanent blemishes in the target surface, by extraordinarily sensitive ("hot") pixels which result in recurrent bright spots at fixed locations, or by radiation-induced events within the UV converter which result in randomly placed, nonrecurrent bright spots (Ponz, 1980 a,b).

Ponz (1980 a,b) has described an algorithm for detecting in raw images bright spots on the basis of their limited spatial extent and unusual brightness values, primarily through a median filtering technique. The IUESIPS program BSPOT provided by VILSPA incorporates this algorithm to flag bright-spots as described below. At GSFC, this program was implemented on 19 November 1982.

Let DN (i,j) be the DN value of the pixel at line i, sample j. Further, let AVE and MED represent operators which return the weighted average and median values of their argument, respectively. Then the pixel at (i,j) is detected as a bright spot if

(3-1) 

DN(i,j) > AVE{DN(k,l)} + Delta

and
(3-2) 
DN(i,j) > MED{mDN(k,l)} + Delta

where Delta is a DN threshold value, and (k,l) are positional elements of a spatial window centered on the pixel at (i,j) and consisting of seven pixels along the image diagonal (i.e., nearly along the dispersion direction). The condition in equation 3-1 is included to reduce, in the general case, the number of times the median operation in equation 3-2 is performed, thereby saving computer time.

In practice, the spatial windows are weighted according to the weights (0, 0, 1, 0, 1, 0, 0), and a threshold value of del = 90 DN is employed. Future studies may examine the feasibility of parameterizing the threshold value according to the image background level, which could improve detectability of bright-spots on very high background images. The area of the image searched for bright-spots corresponds to that containing the spectral orders. Pixel locations detected as bright-spots are written to a disk data set subsequently read by the spectral extraction routines Section (7) so that extracted fluxes derived from bright-spot pixels may be flagged appropriately.

Ponz (1980 a,b) has published partial listings of recurrent bright spots in the SWP and LWR cameras which are reprinted here as Tables 3-1 and 3-2. The table entries include the line and sample positions both in raw and geometrically corrected (see Section 4) frames of reference and the approximate corresponding wavelengths for the various dispersion modes and apertures. The notation "B" means the background, rather than gross, spectrum is generally affected. Ponz estimates the expected error in wavelength for low dispersion to be 5 Å, and for high dispersion to be 0.3 Å. Double high dispersion wavelength entries in certain instances indicate that adjacent orders may be affected.

Imhoff (1984c) has provided positions of additional permanent belmishes in the LWR and LWP cameras, given in Table 3-3.


Table 3-1:   "Hot" Pixels in the SWP Camera
RAW GEOMD Low Dispersion High Dispersion
Large Ap. Small Ap. Large Ap. Small Ap.
Line Sample Line Sample Wavelength (Å)
292 413 295 412     1379.6 B 1378.7 B
1393.6 1392.6
352 501 357 500     1330.2 B  
1343.0 1342.2
392 127 386 123 1795 B   1859.1 1857.8
   
398 521 404 520     1357.9 B 1357.0 B
1371.4 1370.4
410 535 416 534     1358.5 1357.6
1372.0 B 1371.0 B
482 342 481 336     1686.7 1685.6
   
568 387 563 112        
2060.2 2058.9
611 127 613 380     1779.0 B 1778.0 B
1756.0 B 1755.3 B

Table 3-2:   "Hot" Pixels in the LWR Camera
RAW GEOMD Low Dispersion High Dispersion
Large Ap. Small Ap. Large Ap. Small Ap.
Line Sample Line Sample Wavelength (Å)
126 291 120 315       1904.8 B
1919.3 1920.5
170 200 156 222 ~1780 1775 B    
   
175 369 174 394       2153.6 B
2172.5 2173.9
178 610 186 648        
2732.0 2733.8
208 391 207 415     2258.5 B 2282.4 B
2280.0  
215 326 210 348   2130   2117.0 B
2135.3 2136.7
257 323 251 345 ~2190     2199.7 B
2198.2 2178.8
333 317 326 335       2290.3 B
2288.9 B 2268.0 B
412 385 407 401       2543.8 B
2570.2 2572.0 B
434 479 434 498       2786.3 B
2818.7 2820.5 B
518 545 521 563       3086.0
3084.0  
532 307 521 316     2550.8 2552.3
2579.2 B  
680 332 673 335        
2838.0 2839.8

Table 3-3:   Permanent Blemishes in the LWR and LWP Cameras
  Position in Raw Image  
Camera Line Sample Comments
LWR 169 499  
364 60  
LWP 101 525  
205 319  
396 384 Fuzzy patch at lambda ~ 2482 Å in order 93
409 208 Hole at lambda ~ 2880 Å in order 80
426 435  
455 35  


3.3 PARTIAL-READ IMAGE PREPROCESSING

"Partial-read" images are those for which only a portion of the the target has been read. By not reading a full 768 x 768 array, a substantial fraction of the operations overhead time associated with the camera readout and subsequent preparation for next exposure is saved. Partial-read images are used only in the low dispersion mode and are always read out in a standard way so that a camera-dependent rectangular partial image, sufficient to encompass the entire region normally extracted in low dispersion processing (see Sections 5 and 7), is generated. Table 3-3 lists the standard parameters defining the partial-read areas in raw image space for the IUE cameras.

The partial-read images are preprocessed by using the program INSERT to imbed the partial-read area into a full 768 × 768 array for which DN values outside of the partial-read area are zero. This is done to enable the normal IUESIPS processing, which works on 768 x 768 images, to occur without further special consideration of the partial-read nature of the images; see also Section 5.3.

Note that the partial reads are being evaluated to insure that the process has no permanent damaging effects on the cameras. Thus they have not yet been routinely used for GO observations.



Table 3-4:   Standard Partial-Read Parameters
Parameter LWP LWR SWP SWR
Starting Line 99 73 36 135
Starting Sample 31 123 33 175
Number of Lines 528 528 528 480
Number of Samples 576 624 528 576



next previous contents