Scatter plots are primarily used as data visualization tools. Each plot shows the correlation between the histograms for two channels.
Pixel distributions for the two specified image channels display in the scatter plot using one channel as the x-axis and the other as the y-axis. They allow you to see where the majority of data values (or pixels) are concentrated. Frequency values at each point are color coded. Scatter plots also calculate relevant statistics and display at the bottom of the scatter plot.
Natural groupings of the spectral data are best illustrated with a two-channel data set. For image data with more than two channels, it is difficult to plot the values and visually identify natural spectral groupings. Statistical techniques can be used to automatically group an n-dimensional set of observations into natural spectral classes. This procedure is called cluster analysis.
A scatter plot can reduce the number of channels used for a classification. If two channels have a very high correlation, you can omit one or the other as input for the classification. You can also determine which portion of the spectra a given bitmap or training area occupies, and you can use a scatter plot to determine the homogeneity of a bitmap or training area. If the scatter plot for the bitmap is tightly clustered with few outlying pixels, the spectral response for that area is homogenous in the selected image layers.
Plot Scale
For 8 bit imagery, the scatter plot axis are 256 pixels by 256 pixels. The top-left pixel represents the number of pixels with a value of 255 for the input channel on the y-axis and zero for the input channel on the x-axis. The bottom-right pixel represents the number of pixels with a value of 255 for the input channel on the x-axis and zero for the input channel on the y-axis. When images are outside the 0-255 range, the imagery is scaled to fit within that range. When images are outside the 0-255 range, the maximum digital number (DN) value is used for the plot scale.
Pixel Brightness
Is determined by the frequency of pixels in the image with a given gray-level value. Bright areas indicate common combinations and black areas indicate combinations that rarely occur.
Scatter plots typically show a bright smear in one area of the plot. By default, the plot appears with input channel 1 on the x-axis and input channel 2 on the y-axis.
In the controls area of the Scatter Plot window, you can specify the channels you want to show as the X and Y axes. A color scheme for the plot and a lookup table (LUT) can be applied to either channel.
Mask
The Mask option allows you to create a scatter plot of a region under a bitmap mask. You can also create a scatter plot of the entire raster. When creating a scatter plot of the entire raster, you set the Mask option to None. The Mask list box displays all of the bitmap layers that are in the current area. The bitmap layers are listed whether they are saved or not and for each saved bitmap, both the file and layer names are displayed.
Statistics
The Statistics section displays the linear equation derived from a linear regression calculation and the correlation coefficient associated with the scatter plot. A value of 'N/A' (Not Applicable) is given if these statistics cannot be calculated (usually if one of the selected channels is empty). The correlation coefficient measures the similarities of the two image channels. A value of one indicates a complete correlation between two images, whereas a value of zero indicates there is no correlation between images. The A -1 value indicates a negative correlation.
Hyperspectral Metalayer
If you want to view the scatter plot of a hyperspectral metalayer, you must first open the metalayer in the Maps tree.
© PCI Geomatics Enterprises, Inc.®, 2026. All rights reserved.