Special Pixels¶

What are special pixels?¶

Special pixels are defined to distinguish valid pixels from non-valid pixels in ISIS. This is very important to scientists who evaluate the density values, and draw conclusions about the physical properties of a scene.

How are special pixels set?¶

There are two ways that Digital Number (DN) values are set to special pixels in ISIS:

Instrument Special Pixels : These types of special pixels are set during the ingestion of the data into ISIS because the acquired measurement either exceeded the dynamic range of the sensor based on its inherent properties (such as sensitivity) and settings (such as gain) or the measurement wasn't collected due to an instrument or transmission error.
- Low Instrument Saturation (LIS) : The sensor registered a value too low to be measured accurately (i.e. undersaturated).
- High Instrument Saturation (HIS) : The sensor registered a value too high to be measured accurately (i.e. oversaturated).
- No Data Collected (NULL) : The instrument did not collect a measurement due to an sensor malfunction (such as a damaged CCD element) or an error in transmission prevented the value from being recorded.
Representation Special Pixels : These types of special pixels are set due to being processed by ISIS applications that cause the DNs to fall outside the data range of the file, either due to settings defined by the user or the effects of the algorithms within the application.
- Low Representation Saturation (LRS) : The resulting DN calculated by the application fall below the possible range of values (i.e. undersaturated).
- High Representation Saturation (HRS) : The resulting DN calculated by the application are higher than the possible range of values (i.e. oversaturated).
- Data Removed (NULL) : NULL values are set by an application when it removes data, such as during masking or geometric warping.

Special Pixels in ISIS¶

How many special pixel values exist in ISIS?¶

There are five special pixel values defined in the ISIS software system for 16-bit and 32-bit data type. Each of the five special pixels is denoted by a specific value that represents that specific type of invalid data. Having one of these special value is the only way ISIS can tell that a given pixel is in fact a special pixel of a particular type.

The NULL special pixel identifies pixels where either the instrument failed to collect data, or data was removed or empty pixels created during processing.

The LRS (Low Representation Saturation) and HRS (High Representation Saturation) represent saturated pixels that fall outside the minimum and maximum valid range of the file respectively.

The LIS (Low Instrument Saturation) values represent pixels that fall below the lowest value the instrument can accurately record. The HIS (High Instrument Saturation) values represent pixels that exceed the highest value the instrument can accurately record.

See the Special Pixel Demo to visualize the concept of special pixels in a cube.

Non-differentiation of Special Pixels in 8-bit Cubes

8-bit data can only has two special pixel types, represented by values of 0 (NULL, LIS, and LRS) and 255 (HIS and HRS). Thus 8-bit data does not distinguish between NULL, LIS, and LRS pixels, nor between HIS and HRS pixels. You may lose data differentiating special pixels if you convert a higher bit-depth cube down to an 8-bit cube.

Numerical and Visual Representation of Special Pixels¶

Visual refers to the appearance of a special pixel in an ISIS display application, like qview.

Special Pixel Name	32-bit*	16-bit	8-bit	Visual
NULL	-3.40282e+38	-32768	0	black
Low Representation Saturation (LRS)	-3.40282e+38	-32767	0	black
Low Instrument Saturation (LIS)	-3.40282e+38	-32766	0	black
High Instrument Saturation (HIS)	-3.40282e+38	-32765	255	white
High Representation Saturation (HRS)	-3.40282e+38	-32764	255	white

Differentiating 32-bit Special Pixels

The 32-bit values representing different special pixels do differ just like those representing the 16-bit values. However, the variations are on the last digit, and thus are not visible in this display since it does not show all 38 digits.

Are the special pixel values propagated between different bit types?¶

The special pixel values are converted without loss of information between 16-bit and 32-bit data. See the chart above to see what the values are set to after conversion from one bit type to the other. For example, if a pixel in 16-bit data with a DN of -32766 (thus representing an LIS pixel) is converted to 32-bit data, its DN value will be changed to that value representing an LIS pixel in 32-bit format.

If the image is converted from 16-bit or 32-bit to an 8-bit data type, then all the LIS, LRS, and NULL special pixels have their values set to 0, and all HIS and HRS special pixels have their values set to 255. Note that this conversion results in a loss of information, as it is no longer possible to distinguish between LIS, LRS, and NULL pixels since they all have the same pixel value of 0. Similarly, converting from 16-bit or 32-bit to 8-bit will make it impossible to distinguish between HIS and HRS pixels (both having the value 255).

Is the NULL special pixel defined as 0 for all bit types?¶

No. The DN value 0 is defined as NULL (or LRS or LIS) only in an 8-bit file. For the 16-bit and 32-bit data, 0 is a valid DN!

Working with Special Pixels¶

Q: How do I find out if there are any special pixels in my file?

A: Statistics: hist & stats

There are two programs stats and hist that will provide statistics about the pixels of an image.

Output from stats

Group = Results
    From                    = Intro2isis.cub
    Band                    = 1
    Average                 = 61.770642254011
    StandardDeviation       = 8.0295403784653
    Variance                = 64.473518689404
    Median                  = 62.0
    Mode                    = 66.0
    Skew                    = -0.085692730285373
    Minimum                 = 2.0
    Maximum                 = 252.0
    Sum                     = 76029592.0
    TotalPixels             = 1271424
    ValidPixels             = 1230837
    OverValidMaximumPixels  = 0
    UnderValidMinimumPixels = 0
    NullPixels              = 40587
    LisPixels               = 0
    LrsPixels               = 0
    HisPixels               = 0
    HrsPixels               = 0
End_Group

Viewing images in qview¶

You can display the image on a monitor using the program qview, and position the cursor over a very dark pixel. The DN displayed on the screen will either be NULL, LIS, LRS, or a very low DN value. The HIS and HRS pixels will be displayed as white pixels on a monitor.