8. Rescaling

Rescaling is a polar2grid component shared by most backends that rescales gridded image data via the Rescaler object. Rescaling provides simple enhancement capabilities to prepare the product data for an image format. The following sections describe each scaling function in the rescaling code. The defaults provided by Polar2Grid will create nice looking images suitable for most users.

The functions described below are intended for use by polar2grid backends via the Rescaler object, but can be used elsewhere if needed. These functions ignore any fill/invalid values in the data. Common values for an 8-bit (0-255) output are shown as an example. The Rescaler object uses configuration files to set how bands are scaled. The function keyword arguments are passed from the configuration file values.

Note

Rescaling configuration files will change in future versions due to PyTroll collaboration and merging. Please contact us if you have a custom configuration that you need converted to the new (similar) format.

There is a shared clip configuration parameter that defaults to True that will clip data after scaling to the output data limits.

8.1. Linear

method:linear
min_in:Minimum value for linear parameter calculations
max_in:Maximum value for linear parameter calculations
flip:Use an inverse linear calculation on the input data

Note

If min_in and max_in aren’t specified they are calculated on the fly from the provided data.

Note

In the default configuration file this method is used as the default rescaling method

\text{m} = (max\_out - min\_out) / (max\_in - min\_in) \text{b} = min\_out - m * min\_in \text{rescale\_data} = \text{data} * m + b

Example (10-255 from 173-300):

\text{m} = (250.0 - 10.0) / (300.0 - 173.0) = 1.90 \text{b} = 10.0 - 1.90 * 300.0 = -560.0 \text{rescale\_data} = 1.90 * \text{data} + -560.0

8.2. Square Root Enhancement

method:sqrt
min_in:Minimum input value (default: 0.0)
max_in:Maximum input value (default: 1.0)
inner_mult:Multiplier before calling sqrt (default: 100.0 / max_in)
outer_mult:Multiplier after calling sqrt (default: max_out / sqrt(inner_mult * max_in))

\text{rescaled\_data} = \operatorname{round}(\sqrt{\text{data} * inner\_mult} * outer\_mult)

Example (0-255 from 0-1 data):

\text{rescaled\_data} = \operatorname{round}(\sqrt{\text{data} * 100.0} * 25.5)

8.3. Brightness Temperature

method:btemp
threshold:Temperature threshold in Kelvin when to do high or low piecewise operations (default: 176/255.0 * max_out)
min_in:Minimum input value
max_in:Maximum input value

\text{rescaled\_data} = \begin{cases} \text{linear\_scaling}(\text{data}, min\_in, threshold) & \text{temp} < threshold \\ \text{linear\_scaling}(\text{data}, threshold, max\_in) & \text{temp}\ge threshold \end{cases}

Example (0-255 from brightness temperature data):

\text{rescaled\_data} = \begin{cases} 418 - (1 * \text{temp}) & \text{temp} < 242.0 \\ 660 - (2 * \text{temp}) & \text{temp}\ge 242.0 \end{cases}

8.3.1. Fog (Temperature Difference)

method:temperature_difference
min_in:Minimum input value (default: -10.0)
max_in:Maximum input value (default: 10.0)

\text{clip\_min} = min\_out + 5 \text{clip\_max} = 0.8 * (max\_out - min\_out) \text{rescaled\_data} = \text{linear\_scaling}(\text{data}, clip\_min, clip\_max, min\_in, max\_in) \text{rescale\_data}[\text{rescale\_data} < clip\_min] = clip\_min - 1 \text{rescale\_data}[\text{rescale\_data} > clip\_max] = clip\_max + 1

8.4. Inverse Linear

method:unlinear
m:Factor in linear equation
b:Offset in linear equation

\text{rescaled\_data} = (\text{data} - b) / m

Example (0-255 from 0-1 data):

\text{rescaled\_data} = (\text{data} - 0.0) / 0.00392

8.5. Lookup

method:lookup
min_in:Same as Linear scaling
max_in:Same as Linear scaling
table_name:Name of lookup table to use (default: crefl)

Note

The table_name argument is optional. The choices are currently hardcoded in the software. Default is useful for True Color and False Color images.

\text{rescaled\_data} = \text{available\_lookup\_tables}[table\_name][ {linear\_scaling}(\text{data}) ]

8.6. Land Surface Temperature

method:lst

Same as Linear scaling, but 5 is added to min_out and 5 is subtracted from max_out and data is clipped to these new limits after scaling.

8.7. Cloud Top Temperature

method:ctt

Same as linear scaling, but 10 is added to min_out and 5 is subtracted from max_out and data is clipped to these new limits after scaling.

8.8. NDVI

method:ndvi
min_in:Minimum input value (default: -1.0)
max_in:Maximum input value (default: 1.0)
threshold:Threshold between ‘low’ and ‘high’ operations (default: 0.0)
threshold_out:Output maximum for ‘low’ operations and minimum for ‘high’ operations (default: 49 / 255.0 * max_out)

\text{rescaled\_data} = \begin{cases} \text{linear\_scaling}(\text{data}, min\_out, threshold\_out, min\_in, threshold) & \text{data} < threshold \\ \text{linear\_scaling}(\text{data}, threshold\_out, max\_out, threshold, max\_in) & \text{data}\ge threshold \end{cases}

8.9. Passive

method:raw

A passive function to tell the rescaler “don’t do anything”.