EUMETCastView v2.1.9

2.1.9

An Oblique Mercator projection of a pass over the South Pole is no longer torn
in two and squashed into a strip. The Inv boxes on the METimage tab now work
under every histogram method: CLAHE used to ignore them, and None 95% and None
100% showed an inverted Equalize instead of themselves. Saturated VII pixels
show white instead of as a black hole. The VII CLAHE clip limit has a slider of
its own. And "normal size" in the image view is one image
pixel per screen pixel again on a scaled display.

Projections

• An Oblique Mercator pass over the South Pole stays in one piece.
  omerfor gets the distance along the central line from an atan, so it
  jumps by one whole period of the line — about 40 000 km — and with the
  two-point definition the jump sits at the southern end of the line. A pass
  that reached the southern turning point was cut in two, its halves a period
  apart, and the bounding box stretched the canvas round the whole line; a
  pass over the North Pole never came near the jump, which is why those were
  right. The coordinate is now wrapped to within half a period of the middle
  of the pass. Ten VII granules over 81.3° S went from a 40 186 km canvas to a
  4 107 km one; a North Pole pass comes out byte for byte as before. The fix is
  in the projection itself, so VII, VIIRS M and DNB, MERSI and AVHRR all have
  it, and so do the overlays drawn on it.

Metop-SG VII

• Saturated pixels are white, not a black hole. Over a bright enough
  target the solar channels climb past the channel's valid_max onto a
  plateau — the detector saturates — and the reader dropped every count above
  valid_max as no data, which the image leaves transparent. The brightest
  spot in the granule came out black with a white rim, in every solar band.
  Those counts are held at valid_max now, the brightest value the channel
  shows; the fill value, which also lies above it, still marks missing data.
  Bands and recipes both read through the same function, so both are fixed.
  In the granule that showed it, vii_668 went from 2 415 missing pixels to
  none. Granules without saturation compose exactly as before.
• CLAHE follows the Inv boxes. CLAHE is a whole-image pass: it throws away
  the image the segments composed and rebuilds it from the radiances, and that
  rebuild never read the invert flags, so a ticked Inv box left the CLAHE
  image as it was. Each inverted channel is now turned over after the pass.
  Pixels without data stay transparent, as they are under the other methods —
  inverted they would otherwise have come out white.
• None 95% and None 100% invert as themselves. An inverted channel took its
  colour from the equalize table whatever the method, so ticking Inv under
  None 95% or None 100% showed an inverted Equalize. Each method now picks its
  colour first and the Inv box turns that over, the way the projections
  already did it. Inverted Equalize moves by at most one grey level: the old
  path truncated where the plain one rounds, so it was never exactly the plain
  image turned over.
• A slider for the CLAHE clip limit. Under the Histogram matching combo,
  from 1.0 to 10.0 like the geostationary and AVHRR ones, enabled only while
  CLAHE is selected; it re-runs the pass when a drag is released, or at once
  on a click or a key. The limit used to be a fixed 6.9, which is where the
  slider starts, so nothing changes until it is moved. It is kept in the INI
  as clahecliplimitvii, apart from the limit the other two sliders share.

Image view

• Normal size is one image pixel per screen pixel. originalSize only
  reset the view transform, which puts an image pixel on a logical pixel — and
  on a display scaled by 125 % Qt hands the application a device pixel ratio
  of 1.25, so "normal size" was really 125 %. It now divides that ratio back
  out, and the percentage in the title follows, rounded rather than truncated.