DSTRIPE

Description

Parameters

Parameter descriptions

Input: Input raster channel

Specifies the image channel to be destriped.

Cycle

Specifies the number of line in the striping cycle. Although the maximum value is 128, the destriping algorithm performs better when the cycle value is small. A destriping cycle of 10 or less is recommended.

Output: Output raster channel

Specifies the output channel to receive the destriped image. If this parameter is not specified, the input channel is used for output.

Details

DESTRIPE corrects the cyclical striping effects of an image caused by different detector signal response.

Some type of imaging devices (for example, LANDSAT satellites) use arrays of detectors, with each detector recording a line of imagery. Each pass of the detector array creates a number of output lines; complete images are made up of sets of these passes. If the detectors are not properly calibrated, the output image may exhibit an undesirable striped effect, with the stripes being the number of detector lines wide. DSTRIPE can be used to minimize this effect.

The user specifies the image file (FILI), the image input channel (DBIC) that has striping, the number of lines in the striping cycle (CYCLE), and the image output channel (DBOC) to receive the destriped image.

The image analyst must be extremely careful in determining the value for CYCLE. If the specified value does not match the real striping cycle on the image, the destriping algorithm may provide poor results. Typically, if the image is at full resolution from the imaging device, CYCLE would be the same as the number of detectors. If the cycle number is not known beforehand, it is often possible to determine this number by transferring a full-resolution part of the image to the display and using the cursor and cursor position to count lines.

Algorithm

DSTRIPE corrects striping effects of an image channel through multiple passes by modifying its histograms.

An initial pass is made through the striped image. CYCLE number of histograms are built, one for each line in the strip. At the same time, an overall reference histogram is built for the entire image.

At the end of the first pass, the histograms for each CYCLE line are compared to the overall reference histogram, and correction lookup tables are created for each line based on histogram matching.

During a second pass through the striped image, each line is modified using the corresponding correction lookup table, effectively removing the striping.

This algorithm assumes that the histograms have the same probability distribution. This implies that results will be best if the input image used is large (so that variations will average out) and if the cycle is small (so that detectors, as a group, are passing over similar data).

If these assumptions are violated, DSTRIPE may only reduce the striping effects instead of removing them. It is also possible that striping may be introduced into other parts of the the image.

References

Horn, B. K. P., and R. J. Woodham, (1979). Destriping Landsat MSS images by histogram modification. Computer Graphics and Image Processing. Vol. 10, pp. 69-83.

Poros, D. J., and C. J. Peterson, (1985). Methods for destriping Landsat TM images. Photogrammetric Engineering and Remote Sensing Vol.51, No. 9, pp. 1371-1378.