2.6 Using histograms to interpret data

Histogram binning Interpreting multiple peaks Histograms from subareas

TUTORIAL

2. Histograms

Image:

ATS_TOA_1C_ 20040201~.N1 Description

Download image (7.7 MB)

Useful information:

Image data: A grid of pixels

In this section we will look at how histograms may help to interpret data from remote sensing scenes and distinguish between diferent categories of pixels. For this we will use the AATSR scene from south western Africa.

Activate (or open) the btemp_nadir_1100 band of the image ATS_TOA_1COLRA20040201_082356_000001252023_00478_10051_1629.N1
Select the whole image ( [CTRL+A] ) and create a new histogram as described in section 2.1

Histogram binning

Move your cursor along the image and read the frequency information from the status bar as described in section 2.3. You should see two different data values for each cursor position - one a little higher than the other. These represent the upper and lower end of the numerical range represented by a particular bar in the histogram.
This is different from the histogram representation of the 8-bit GIF image from January 1997, where each bare represented only one data value. This grouping of the data values is known as binning.

Question 1
a) How many different integer values are combined in each bin in this histogram?
b) Can you think of a reason why Bilko bins this 16-bit data, but not the 8-bit data?

Interpreting multiple peaks

You will have noticed that the histogram of this image has four fairly distinct peaks - a low, broad peak at low data values, two tall narrow peaks close together near the centre of the frequency distribution, and another low, broad peak at the high temperature end of the frequency distribution.

Such clearly distinguishable peaks often indicate that the image contains pixels from different categories or surface types, each with its only characteristic frequency distribution. These may sometimes overlap, but there is frequently a local minimum - a 'valley' of low frequency that indicates the boundary between the different pixel types.

Activity / Question 2
a)	What are the pixel categories represented in this image? Can that explain the number of peaks you see?
b)	Use your mouse and arrow keys to move the cursor along the x-axis and find the digital value corresponding to the maximum pixel number for each of the four peaks. What brightness temperatures do these values correspond to?
c)	Approximately what temperature ranges are covered by the four peaks, and what type of surface are they likely to represent? (If you can't be precise about the range covered by a peak don't worry - there are often good reasons why the peaks might overlap.)

Histograms from sub-areas

One way to check your assumptions about the four peaks in Answer 2, would be to look for cloud-free areas of land and sea in the 0550 band of the image, where the clouds are distinctly visible because of their high reflectance. A histogram based on such a single-category area should correspond broadly to a peak representing this type of surface in the histogram of the whole image.

Activate or open the 550 band, and use box selection with the cursor to draw a box around an area of ocean that does not appears cloud free.
Open the Go-to dialog ( [CTRL+G] ), uncheck the Coords box to see the pixel coordinates rather than the geographical positions, and and make a note of the x, y, DX and DY values. (If you are unsure where to place your box selection the area x=110, y=420, DX=90, DY=180 is a good
Activate the image of the 1100 band and use the Go-to dialog to select the same region in this image.
With this area selected open a new histogram document ( [ CTRL+N] >Histograms ).

As you can see the new histogram has only a relatively narrow peak. Depending on where you mad your selection its position will vary slightly; in the example above, the frequency distribution of the selection peaks at 28800, or 288°K (15°C).

Answers:
(Resizable
pop-ups)

Answer 1

Answer 2

Answer 3

Back up to:
Q1 Q2

Activity / Question 3
Use the 0550 image to select a box containing both land and sea, but no cloud (for example x=120, y=420, DX=390, DY=180), and create a new histogram from the box selection.
a)	How many peaks does this histogram contain? Which peak is missing?
b)	Explain the differences between the histograms.

In the next section you will need to refer to the histogram you just created during Activity 3, so save it in a convenient place, giving it a name such as ATS_TOA_1COLRA20040201_082356_ocean_land.hst (a long name, but it will keep the histogram and the original image together in your file lists).

Leave this histogram, the ATS_TOA_1COLRA20040201~.N1 file structure window, and the btemp_nadir_100 image window open, but close all other windows before continuing.

Next: Histograms and 'Redisplay'