next previous
Up: An image database


Subsections

2 DSS material

2.1 Description of DSS scans

The material comes from the set of 102 compressed CD-ROM's of the Digitized Sky Survey produced at the Space Telescope Science Institute. Each CD-ROM contains the digitization files of several plates. The Northern hemisphere ( $\delta \geq +3 \deg$) contains 644 plates, noted xe001 to xe643 and one additional (xe1001). The Southern hemisphere contains 894 plates noted s001 to s894. Each plate is constituted of 784 ( $28 \times 28$) individual scans of $500 \times
500$ pixels[*]. Each individual scan is labelled with a 2-symbol extension (from 0 to r), e.g., s828.00 to s828.rr. The scan *.11 is the most South-Eastern one, while the scan *.rr is the North-West one. We skipped all scans labelled with 1 or r (i.e., *.1x, *.x1, *.rx, *.xr) in order to reject the extreme edge of each plate. Thus, we processed about one million elementary scans. The size of a pixel is 1.70 $^{\prime\prime}$. Each elementary scan is about $14.2\hbox {$^\prime $ }\times 14.2\hbox {$^\prime $ }$. The Northern part is digitized from E(red) plates, while the Southern part comes from IIIa-J plates. An example of a typical elementary scan is given in Fig. 1.


  \begin{figure}
\par\includegraphics[width=8cm]{ds1851f1.eps}\end{figure} Figure 1: A typical image (s828.ap) from the DSS CD-ROM's. This image is only $12.6\times 12.6$ mm on the original plates. The frame is $14.2\hbox {$^\prime $ }\times 14.2\hbox {$^\prime $ }$. North is on the bottom side, East is on the left side. The seven galaxies clearly visible on the scan are new ones which were recognized by the automatic program

2.2 Source extraction

The source extraction is made from uncompressed scans in the same way as in our previous papers (Paturel et al. 1996; Vauglin et al. 1999). The sky background is assumed to be homogeneous over an individual scan. Its mean intensity $I_{\rm bg}$ is calculated from the maximum of the histogram of pixel intensities. The standard deviation $\sigma_{\rm bg}$ is calculated by symmetrizing the low intensity side of the histogram. The threshold for source extraction is chosen as $I_{\rm bg}+3 \sigma_{\rm bg}$. Only the sources having more than 36 pixels are kept. This means that the smallest objects have a size of $\sqrt{37}$ pixels, i.e., $10\hbox{$^{\prime\prime}$ }$. Further we impose that the number of pixels on one side of the matrix of pixels is larger than or equal to three. If we note npx, the number of pixels per line and nli the number of lines, this means that $npx \geq 3$, $nli \geq 3$ and npx.nli > 36.

2.3 Equatorial coordinates

For each object we calculate the $\overline x - \overline y$ mean position of the matrix. The mean is obtained by weighting each pixel with its intensity. The J2000 equatorial coordinates are then calculated using the plate solution calculated at the Space Telescope Science Institute. The coefficients of the 13-th order polynomial solution are read in the header file associated which each plate. The internal accuracy is about 3 $^{\prime\prime}$ in right ascension and declination, near the equatorial plane. Near the Northern pole the accuracy is about 4 $^{\prime\prime}$. Near the Southern pole it is about 5 $^{\prime\prime}$. This is in agreement with the results found by a recent study of coordinate accuracy (Paturel et al. 1999; Paturel & Petit 1999). Most of the uncertainty comes from the positioning of the galaxy center.

2.4 Cleaning of sources

A treatment is then applied on matrices corresponding to overlapping objects in order to separate them into their different components. The basic assumption in this treatment is that astronomical objects have a central symmetry.

First of all, we determined the number of maxima in the pixel matrix. This is done as follows: when a maximum is found, the region around its position is inhibited and the following maxima are looked for, outside this region. The inhibited region around each maximum is actually the ellipse centered on the considered maximum and tangent to the nearest edges of the matrix.

The decomposition of a matrix in several matrices is then made in the following way: all pixels of the original matrix are considered one after the other. For a given pixel P(i,j) we are searching for its symmetrical counterparts with respect to the n maxima $M_k \ (k=1,n)$. If the symmetric counterpart of a given pixel P(i,j), calculated with respect to the k-th maximum, is outside the matrix, or if it has an intensity below the sky background, the given pixel P(i,j) is not attributed to the object re-constructed around this k-th maximum. If the pixel P(i,j)belongs to several re-constructed objects, the pixel intensity is simply shared with equal weight between each object. No attempt has been made to share this intensity in a more refined way because the pixel intensities are not additive. The objects which result from this decomposition always have a central symmetry in accordance with our basic assumption. In the final catalogue a flag will remind us a given object results from such a decomposition process. The matrices which are truncated by the edge of the scan are also extrapolated by symmetry if the maximum itself is not on the edge.


next previous
Up: An image database

Copyright The European Southern Observatory (ESO)