Terapix Star Formation Region IC 1396, © 2002 CFHT
T0004 explanatory table of Deep+Wide stacks
by YME YME - Updated November 13th, 2008

Back to Terapix T0004 page.

This table is valid for the T0004 release only. Please do not use it as an explanatory table of T0003 or any previous releases. We strongly recommend CFHTLS registered users to read carefully this page when using the Synoptic T0004 Table.

Stacked Images All stacks are produced by a median of an input image list, weighted accordingly, and combined using SWarp, with a Lanczos3 interpolation kernel. Equatorial Coordinate system, J2000 (FK5). Projection: TAN.

The SWarp configuration file contains CENTER , IMAGE_SIZE , PIXEL_SCALE parameters that will be used by all filters of a given field. For T0004, as for previous releases, all images have 19354x19354 pixels, 0.186 arc-second/pixel. The center positions of each stack are fixed and correspond to the center positions of each CFHTLS pointing, as pre-defined in the CFHTLS Deep and Wide center list. The "Cartesian" W(i,j) tile organisation and name definition of Wide target positions are defined in the W1 target list, the W2 target list, the W3 target list and in the W4 target list.

Astrometric reference catalog used: 2MASS. Each image is composed of one single FITS file (no more MEF with 36 FITS files; CCDs are no longer singled out after stacking).

FILENAME: CFHTLS_[D*/W]_[filter]_[RA][DEC]_T0004.fits , where the ''D*'' stands for Deep D-25 (25% best seeing images) or D-85 (85% best seeing images) data and ''W'' for all Wide fields [Col. 4-8 of the Synoptic T0004 table (labeled as "Im")].

A list of all individual images included in each stack is given in the Qfits-out output table. Each image listed in this table is then linked to its Qfits-in, so that one can can inspect all assessments done on all images of a stack.

Important new feature for T0004: Images (as well other compressed files at CADC) are compressed using gzip (.gz files) instead of the RICE algorithm as in previous releases.

Weight map images FITS weight map image. There is one weight map image for each FITS image. Input weight maps are generated during the QualityFITS-in process and output weight maps are produced during the image stacking using Weightwatcher in SWarp. Weight is defined as the local sky background flux variance. The input weight-maps are generated using the master flat field and mask FITS images delivered by CFHT. When used during co-addition, the weighting scheme takes into account the image-to-image flux rescaling, as described in the SWarp User's manual.

FILENAME: CFHTLS_[D*/W]_[filter]_[RA][DEC]_T0004_weight.fits . [Col. 4-8 of the Synoptic T0004 Table (labeled as "W")]

Chi2 Image Constructed by combining images using "SWarp" (hereafter=filterlist)- Pixel values are actually sqrt(chi2) , not chi2 . There is one Chi2 image per field (using joined g, r and i images only), but ONLY when g, r and i stacks are available. Chi2 image is generated using the swarp.conf of the co-added images, but with the 'COMBINE_TYPE' chi2' option.

FILENAME: CFHTLS_[D*/W]_gri_[RA][DEC]_T0004.fits . [Col. 9 of Synoptic T0004 Table]

Catalogs There are two types of catalogues: the '.ldac' and the '.cat'. For the .cat of catalogues, T0004 has u,g,r,i and z , separately, but all chi2-based, and a merged but smaller ugrzi.

-  The ".ldac" catalogues are FITS table. They are produced by SExtractor using only a single image or a stacked image and its weight map. There is one '.ldac' per stack produced during the Qfits-out process. Note that there is also one '.ldac' produced during the Qfits-in process of each image transferred at Terapix from CADC. All ".ldac" derived from a stack process refer to a given filter only, regardless other filters related to the same field (in particular there is no use of chi2 image and no cross-identification). These catalogues are not corrected for galactic extinction. The .ldac catalogues are available via the Synoptic T0004 Table in the same columns as the stack and weight images of each stack [Col. 4-8 of the Synopic T004 Table, (labeled as "L")].

-  The '.cat' catalogues are ASCII files. They are produced only when at least the g, r and i stacks and their ".ldac" are produced. In that case there is one ".cat" for each filter, but they are extracted with the same chisquared detection image and they contain the same objects (i.e., they are all aperture-matched). The '.cat' file contains also the E(b-v) in the last column. Each catalogue id for each filter for a given field refers to a different object. For almost all purposes, the Kron-like pseudo-total magnitudes gives the best estimate of the object's magnitude (in non-crowded regions). Columns of catalogs with both the '.cat' or the '.ldac' extentions are defined in the '.param' files. They are listed in the CFHTLS catalogue parameter list. For example, the Kron-like pseudo-total magnitudes is listed in column (60).

-  Note that in addition, a 'merged' '.cat' catalogue is also produced that merges the '.cat' into one single file but with less information.

FILENAME : CFHTLS_[D*/W]_[FILTER]_[RA]-[DEC]_T0004.ldac or .cat (see below) [Col. 4-8 and Col. 10 and 12 of Synoptic T0004 Table]

Catalogs-chi2 These catalogues were produced using SExtractor in dual-image mode, using the gri chisquared image as reference, as well as the weight-map for the final stack. Each catalogue id for each filter for a given field refers to the same object.

Columns of catalogs with '.cat' extentions are defined in the '.param' files. They are listed in the CFHTLS catalogue parameter list. Note that these catalogues have not been corrected for galactic extinction. However, for stacks having at least gri data, each u, g, r,i, z catalogue contains the value of the galactic E(B-V) extinction, as calculated using the Schlegel et al dust map at the object's position (last column added).

For almost all purposes, the Kron-like pseudo-total magnitudes give the best estimate of the object's magnitude (in non-crowded regions). Users who require z- or u- limited catalogues should use these files with caution.

Important notice: The .cat catalogues are produced using specific selection criteria during the SExtractor step applied to the chi2 image and during the u/g/r/i/z merging phase. There are certainly not suitable for all scientific goals. Please have a look at the configuration and parameter list files. In case you need different selections, you may need to produce your own .cat catalogue, or possibly your own chi2 image.

FILENAME: CFHTLS_[D*/W]_[FILTER]_[RA]-[DEC]_T0004.cat [Col. 10 of Synoptic T0004 Table]

Mask files Built automatically using the "polygon" option of the DS9 visualisation tool to create a "*.reg" file which describes, in world co-ordinates (WCS), the regions of each image which should not be used for scientific analysis. The files mask all bright stars and the boundaries of each field. The size of the masked areas at star position is scaled according to their USNO-B B magnitude. Each mask can be used with DS9 (or SAOimage, after they have been transformed to pixel co-ordinates). Polygons may be added or removed according to user wishes It is important that any users check the masks and modify them according to its scientific needs.

Note that in contrast with previous releases, masks are built with USNO-B data only and are no loner constructed using the i-band images.

FILENAME: CFHTLS_[D*/W]_[RA][DEC]_T0004.reg (note: there is no more filter in the mask name). [Col. 11 of Synoptic T0004 Table]

Merged chi2 catalogues The merged catalogues use the catalogue-chi2 to produce a simpler ugriz catalogue by merging their most useful columns into one single file. Each merged catalogue contains the following information (With the exception of the TERAPIX quality flag, all this information is available in each of the separate catalogues produced from the chi2 image. )

-  (1) id : the SExtractor object id. This corresponds to the object ID which appears in the chi2 catalogue - (2,3) x,y (pixels): object pixel co-ordinates. Pixel scale is 0.186"/pixel - (4,5) RA,DEC; right ascension and declination in J2000 co-ordinates - (6) r2 (pixels): radius enclosing half the light - (7) flag : TERAPIX object flag. Objects with flag 0 are 1 are galaxies or (non-saturated) stars, respectively, outside the masked zones. Objects inside a mask is flaged 0+4 or 1+4 (Note that star-galaxy separation is only carried out on the i* image to a limit of i*=21.0; fainter than this limit, all objects are identified as stars). - (8,9,10,11,12) u, g, r, i, z : Object MAG_AUTO magnitudes in the MEGACAM instrumental reference frame - (13,14,15,16,17) uerr,gerr, rerr, ierr, zerr: Object magnitude errors. - (18) E(B-V): The value of the galactic E(B-V) extinction calculated using the Schlegel et al dust map at the object's position.

-  When MAG_AUTO and its error are missing, these values are set to -99

Relevant input files:
-  u,g,r,i,z chi2-catalogs or only g,r,i chi2-catalogs -> *.cat
-  catalogues are produced with the CFHTLS catalogue parameter listSExtractor parameter list.

-  Masks are the current T0004 .reg DS9 compliant mask files available at CADC

IMPORTANT notice: some u-band data show an offset of 0.20 mag. with respect to expectations. This offset is visible from the stellar colour-colour plots as well as on the comparison with SDSS, when available. It concerns fields obtained after Feb. 2006 and is likely related to a calibration problem at CFHT. In order to correct for this offset we recommend you look at the Obs. Date of u-band data given in Quick T0004 stack summary table

FILENAME : CFHTLS_[D*/W]_ugriz_[RA]_[DEC]_T0004.cat [Col. 12 of Synoptic T0004 Table]

Galaxy counts Galaxy number counts per 0.5 mag., per deg2, computed. This is a rough count estimate based on the single filter detection (no chi2 image used at this stage): no mask, no galactic extinction applied (note that based on Terapix past analyses of T0002 and T0003, using the chi2 image and the mask should push the magnitude limit 1.0 to 1.5 magnitude deeper). Magnitudes are SExtractor MAG_AUTO. No extinction correction has been applied. The magnitude system is instrumental AB ; there is one plot per filter. To each plot a comparison to the galaxy count expectation is overplot, based on previous Megacam CFHTLS Deep galaxy counts.

FILENAME: CFHTLS_[D*/W]_[FILTER]_RA-DEC_T0004_gal_histo.ps. [Col. 13 of Synoptic T0004 Table]

Stellar colour-colour Stars were selected by automatically locating the vertical stellar branch in the compactness(rh)-magnitude diagram in a series of 10 arcminutes cells distributed over each Megacam I-stack; only stars with MAG_AUTO magnitudes in the range [16.5-21.0] mag. have been selected. Magnitude system is instrumental AB.

Stellar tracks from Pickles overploted for comparison. The Pickles input data have not been corrected from atmospheric transmission, nor from galactic extinction. The selected stars of Megacam fields have not been corrected from galactic extinction.

Terapix provide 1 plot per field per filter-triplet and a maximum of 3 plots when the 5 filters are available. A compilation of all colour-colour Deep or all colour-colour Wide plots of T0004 is available via the Synoptic T0004 Table (two merged ps file , last row of the table)

FILENAME: CFHTLS_[D*/W]_RA-DEC_T0004.[FILTER-1]m[FILTER-2]_[FILTER-2]m[FILTER-3].ps, where ''m'' stands for ''minus'' for the colour term (e.g.: rmi = (r-i) = [FILTER-1=r]m(inus)[FILTER-2=i]). [Col. 14 of Synoptic T0004 Table].

Three colour images Produced by combining the stacked FITS images in 3 filters (the 3-filterlist, like irg, igu or zir) using STIFF . These 2 x 2 binned colour images are not optimised for public outreach posters. The R,G,B color balance and cuts are primarily optimised for visual inspection as a quality assessment tool. It is primarily used to check filter-to-filter astrometric calibration and whether colour gradients are visible in objects, resulting from poor overlaps between each filter, in particular far from the centre of the field.

FILENAME: CFHTLS_[D*/W]_[3-filterlist]_[RA][DEC]_T0004.jpg . [Col. 15 of Synoptic T0004 Table].

Mag. ZP Photometric calibration is carried out in two steps.

-  In the first step, the photometric solution provided in the header of each image by CFHT (see CFHT photometric calibration page ) is used to re-scale the zero-point of each image to the values given in the Magnitude ZP columns of the summary table (for a one second exposure at zero airmass).

-  In the second step, overlapping sources on all exposures in a given field for a given filter are compared. All other images are then scaled to the reference images having the CFHT "P" photometric label. A photometric rescaling value is then computed for all image and is expressed as the Mag. ZP_corr term found in the .ahead file. The second step is done by SCAMP. This absolute calibration is accurate to 0.05 magnitudes, stack-to-stack, based on consideration of stellar colour-colour plots and number counts (see below).

The magnitudes of objects in the final stacks are computed as follows: m=30-2.5log(counts). In some fields, offsets (probably relating to photometric calibration problems at CFHT) may need to be applied. See the table the CFHTLS(ZP) / SDSS offset table values for more information. In this case the effective zeropoint is (30+Corr). Note, however, all of the catalogues delivered to CADC were computed with a photometric zero point of 30. The magnitude system is instrumental AB

Dust : galactic extinction The E(B-V) is provided directly as a parameter in the final catalogues. It is derived for each object position from the E(B-V) images from Schlegel et al .

The E(B-V) value derived at an object position is computed using dust.c provided with the Schlegel et al. dust data, with the interp flag set to "y" to return a linearly interpolated value from the 4 nearest pixels.

Astrometric and field-to-field photometric calibrations T0004 astrometry and field to field photometric rescaling were calibrated with the new Terapix tool SCAMP (Bertin et al 2007). As SWarp and SExtractor, SCAMP uses plain text configuration files and can be installed on any POSIX-compliant platform.

SCAMP takes as input SExtractor catalogues and produces as output WCS images headers (.head or .ahead files) that can be read by SWarp to re-scale, re-sample and co-add images. The first task for SCAMP is to examine the image headers and group exposures by sky area, "astrometric context" and "photometric context". An "astrometric context" defines a set of exposures that have been observed with a common, mechanically stable configuration. For the MegaPrime/Megacam CFHTLS data, these are exposures which share the same filter and QRunID. There is one astrometric solution per astrometric context. Likewise, a "photometric context" characterises a set of exposures with the same spectral response, that is, with the same filter. Before proceeding further, the CCDs is each available astrometric context are re-positioned to a common "median" focal plane, to correct to strong CCD-to-CCD jittering introduced by the Elixir astrometric solutions. A subset of the reference astrometric catalogue (2MASS J-band for T0004) surrounding each identified sky area is automatically downloaded from CDS. Cross-correlation is used to match each exposure with respect to the reference catalogue: pixel scales, frame position angles, and shifts in RA-DEC are determined within around 1 arc-second.

Pattern matching is followed by a cross-identification with a search radius of 2" between all overlapping detections. In the context of the CFHTLS, SCAMP computes a polynomial model of Megacam image distortions by minimising a weighted, quadratic sum of differences in position between both overlapping detections and matched sources from the 2MASS astrometric reference catalogue. Residual position errors provide an approximate quality of the final astrometric solution. Roughly, on large scale (larger than a typical dithering step) errors are dominated by those of the reference catalogue which is typically a few tenth of an arc-second. At the smallest scale, the relative astrometric accuracy is typically 17 mas RMS.

Specific T0004 astrometric and photometric calibration details are given here . Note that for T0004, all positions are expressed in the J2000 system and ignore proper motions. Also, the *.head files used for each Megacam image used in co-additions are provided via the Synoptic T0004 Table as a tar file through the 'scamp' link.

A photometric calibration correction is applied to images, prior to image co-addition, in a similar way to the astrometric distortion coefficient determination, by minimizing the weighted quadratic sum of MAG_AUTO magnitude differences from overlapping detections. MAG_AUTO was chosen as it is fairly robust and insensitive to the seeing or the size of sources. Adding these photometric constraints the residual Zero-Point discrepancy seen amongst overlapping "photometric" (that is labeled as photometric "P" from CFHT analysis) fields is about 1-2%. The resulting residual on bright objects (S/N > 100) are typically 0.02-0.05 magnitudes RMS, but the relative ZP uncertainties are much smaller. These ZP uncertainties are dominated by the effects of CCD-to-CCD variations, seeing changes and residual gradients within a Megacam field.

The SCAMP output headers contain astrometric and photometric calibration data and a series of quality assessment output files. The .head and .ahead calibration data are ASCII files that contain a multi-FITS header with keywords providing updated WCS astrometric data and photometric rescaling coefficients attached to each CCD. The photometric rescaling values are the .ahead file located in the same tar file as the astrometric solution (see previous paragraph).

For T0004 astrometric/photometric calibration and rescaling were done with SCAMP 1.3.7-MP. Quality assessments are summarized in the series of plots and tables. Both are available via the Synoptic T0004 Table (restricted access to CFHTLS registered users) and then the scamp.xml file [Col. 16-20 of Synoptic T0004 Table].

FILENAME: for input CFHT/Qfits-in images/catalogues #cfhtnumber-p.fits/ldac , the following output files are produced : #cfhtnumber-p.head and #cfhtnumber-p.ahead

T0004 SCAMP plots and quality assessment data The series of SCAMP plots produced for the astrometric calibration and the field-to-field photometric rescaling are all provided in the Synotpic T0004 Table, in the columns ''Nb of files'', using the ''scamp'' link. One can then get the following:

-  A link to a series of tables, the 'scamp diagnostic outputs' link, that provide detailed information on the input and output data and meta-data: image.ldac file used, pixel scale, RA, DEC, S/N, ZP_corr applied to each image, liste of astrometric and photometric groups, internal and external astrometric and photometric rms error, etc...

-  A tar .tgz file that contains ALL information relevant to the SCAMP processes (plots, tables). The tar contains all .head and .ahead calibration files computed by SCAMP (see next description item), so that anyone can use them to run it own Swarp and get its own series of stacks.

-  A series of png files that provide global views of SCAMP results:

. The 2 chi2/dof values of the best astrometric solutions found for all images calibrated together, for both rows and columns of Megacam images.

. 5 plots that display the internal error (x-internal along the x-direction, x-internal along the y-direction, y-internal along the x-direction, y-internal along the y-direction and a x-y internal error in polar coordinates). Internal errors are derived from the cross-identifications and the position differences of objects in the overlapping regions of each Megacam field.

. 5 plots similar to the previous ones but for the external errors. They are derived from the cross-identifications and position differences of common stars between the relevant Megacam fields and the 2MASS J-band catalogue.

. A series of focal plane distortion maps: one per QrunID and one per filter.

. A series of global views of all Megacam fields projected on the sky. There is usually one per group, but for T0004 only one per Wide field (W1/W2/W3/W4), one for Deep-25 and one for D-85 has been produced.

. A series of plots showing the internal photometric errors along the x and the y axis. There are 2 plots per photometric instrument. As for the internal astrometric solution, photometric internal errors are derived from the cross-identifications and the position differences of objects in the overlapping regions of each Megacam field..

Important Notice about T0004 u-band data For all stacks, whenever possible, the photometric catalogues have been compared to SDSS. A summary table is given is the Table of offsets between ZP(CFHT) and ZP(SDSS). It is important to notice that there is a 0.2 mag. offset offset for some u-band data. It only concerns data obtained after Februray 2006 and is related to a CFHT u-band calibration problem. The 0.2 mag. offset in u-band applies to ALL u-data obtained since March 2006. In particular all W4-u band must be corrected. Before March 2006, no offset correction is needed. For the u-band data sets we do not have SDSS comparison, we therefore recommend to apply an 0.20 mag. offset to stacks done with images having Obs. Date after February 2006. The u-band image Observation Date used for all u-band stacks are given in the last column of the T0004 quick summary table
.head files ASCII files containing the FITS keywords derived from the astrometric solution found by SCAMP and the FSCALE FITS keyword derived from the photometric rescaling using SCAMP as well (that has to be used in the swarp.conf file : FSCALE_KEYWORD FLXSCALE).

All T0004 .head and .ahead calibration data can be obtained from the Synoptic T0004 Table, via the 'scamp' link to a tar .tgz file. The .head files of each image contain a multi-FITS header with keywords providing updated WCS astrometric data and photometric rescaling coefficients attached to each CCD. The photometric rescaling values are the .ahead file located in the same tar file as the astrometric solution (see previous paragraph).

The .head files can be used to re-SWarp and re-stack FITS images, if needed, without need to run the astrometry software again.

Qfits-in Provides a weight map image, the primary '.ldac' catalogue of objects (used later by SCAMP for the astrometric calibration) and a series of quality assessments produced by the QualityFITS tool for ALL images transferred on Terapix disk (STEP1 products of each release). Input images are then installed on the Terapix processing storage system when the release STEP1 process starts. Output weight map images and Qfits-in output are sent to CADC. The Qfits(in/out) tools use SExtractor, Weightwatcher, SCAMP, SWARP and STIFF and a series of more specific software.

The list of images selected for the T0004 release uses the Qfits-in evaluation as well as the following criteria to produce the astrometric .head calibration files : acquisition between June 1st, 2003 and Oct. 24, 2006. ; seeing <1.3 (except u band <1.4); airmass < 1.7; exposure time >180 sec. and Terapix quality class A or B (all CFHT grade 4 and 5 images have been graded C by Terapix and are therefore no longer considered during STEP2). A summary table of selected targets can be saved using the option "Export the whole results in CSV" (ASCII file).

Qfits panels contain 500 x 550 pixels PNG images of each field and their attached weight and background maps, a PSF map image of the field (one PSF model per CCD), histograms of sky background and seeing over each CCD, a preliminary estimate of galaxy and stellar number counts (rough galaxy-star separation, no mask used to avoid defects and very noisy parts of the CCDs in order to use the effective Megacam area. Have a look to the plots of galaxy counts given above that have been done carefully.), with the selected objects shown on rh-mag. diagrams, a summary table (global or for each CCD) with a link to binned 4 x 4 PNG images of each CCD. The PNG field image and its attached weight map are clickable PNG files that provides a 1000 x 1000 resolution PNG image, if needed.

All qualityFITS-in data products that are used in T0004, including individual weight map images attached to each Megacam image, are available at CADC.

FILENAME: for input CFHT images #cfhtnumber-p.fits (CFHT) --> #cfhtnumber-p.ldac (catalogue) , and #cfhtnumber-p_weight.fits (weight map image)

Qfits-out QualityFITS quality assessments information on final stacked image. Provides similar panels with the same information as Qfits-in described above (except for the PSF map image of the field which has a 10' resolution).

The number of images stacked, the total exposure time and the median seeing are automatically included in the Qfits-out page. A table lists also the individual images used in each stack and links each of them to its Qfits-in page.

All Qfits-out were produced with QualityFITS v.1.13.2 . New features have been added as compared to T0003, like 5 full resolution 512x512 sub-images that show the image quality and details at the four corners and the center of each stack (these features are also added to very last Qfits-in outputs). A summary FITS table provides information on stacked images .

All QFITS-out data are available from the synotpic T0004 table [Col. 21-25 of Synoptic T0004 Table]

Seeing The "Terapix seeing" is defined as twice the median FLUX_RADIUS of a selection of point sources on each CCD. FLUX_RADIUS, as measured by SExtractor, is the radius of the disk that contains 50% of the total flux (it is sometimes called rh in the literature).

For a Gaussian profile, the Terapix seeing is equal to the Full Width at Half Maximum (FWHM). For a typical MEGACAM Point Spread Function (PSF), the Terapix seeing is slightly larger ( 10%) than the true PSF FWHM. [Col. 26-30 of Synoptic T0004 Table]

Completeness A rough completeness estimation is computed in Qfits-out, for all stacks of the release. It is derived by adding simulated stellar objects with a Moffat profile and flux spread over the magnitude range relevant for the exposure on image sub-sections extracted near the center of each stack. Simulated stars have an equivalent FHWM of 1".

The completeness plots show, on the x-axis, the magnitude of the simulated stars and , on the y-axis, the fraction of stars which are recovered for each 0.5 magnitude slice. There is one plot per field. Magnitudes are SExtractor's MAG_AUTO. MAG_AUTO are Kron-like "quasi-total" magnitudes. The typical shift with respect to true "total" magnitudes is 0.06 mag (the flux being slightly underestimated).

The 50% and 80% completeness limits are derived from an automated curve fitting using an ad hoc Atanh function having its inflexion point forced to y=50%. Since completeness curve should drop quickly when approaching the limit, the 50% and 80% value range in steepest part of the curve, so the mean limiting 50% and 80% magnitude errors are large, about 0.2 mag. [Col. 31-35 of Synoptic T0004 Table]

Effective FOV Effective area is derived by subtracting from the MEGACAM fields all the masked POLYGON areas defined using the CFHTLS_[D*/W]_[RA][DEC]_T0004.reg mask files delivered to CFHTLS users. Each '.reg' file is then an input mask to Weightwatcher (see the latest news on the news about Weightwatcher). Overlaps between polygons are taken into account so that no area is subtracted several times. The effective area is same for all filters of a given field, this is the reason why the masks file names are the same for all filters. The effective areas of all T0004 stacks are given in the Synoptic T0004 Table. If necessary, more precise (4 digits) effective area values can be found in column (7) of the T0004 stack summary table [Col. 36 of Synoptic T0004 Table].

Note that any users can modify the mask according to his own wishes, but the effective area should then be recomputed

Data processing A detailed description of Terapix release and processing is given T0002 explanatory document . An updated version for the T0004 is under way.

Back to Terapix T0004 page.

Site Map  -   -  Contact
© Terapix 2003-2011