Recursive median filtering

DETAILED DESCRIPTION My copending patent application Ser.
No.
07/173,859 filed March 15, 1988 titled "Automatic Target Detection and Recognition" (hereby incorporated by reference) covers a target screener using a segment labeler/extractor with data from an electronic warfare sensor such as FLIR (Forward Looking InfraRed) imagery; in a Low Altitude Navigation and Targeting InfraRed for Night (LANTIRN) system.
Filtering using the present invention may be used in conjunction with that system to provide data with an improved signal to noise ratio.
RECURSIVE MEDIAN FILTERING Median filtering is a common image signal processing filtering technique.
Its non algebraic basis can provide a better filtering result under many conditions.
Algebraic techniques such as linearly combining adjacent samples results in loss of high frequency signal content.
Median filtering does not combine values but rather attempts to logically select the best data sample among the samples available and therefore more faithfully recreates the original signal.
The beneficial effects of median filtering are further enhanced in this invention by configuring the filter to permit a simultaneous use of both median and algebraic filtering.
The first of these architectures is called recursive median filtering.
A median filter and a recursive median filter block diagram are shown in FIGS.
1a and 1b.
The 3X median filter of FIG.
1a merely always selects the middle value from the three nearest neighbors and is a common filtering technique in the prior art.
To explain the notation used in the description, consider the form of digital data using pulse amplitude modulation for example.
With voice signals the waveform is sampled at fixed intervals with a sampling rate equal to at least twice the highest frequency to be transmitted.
The amplitude of each sample is then encoded in a binary code, usually with some form of pulse compression in accordance with a given law.
For imaging data, a camera or other input device captures the image a frame at a time at fixed frame intervals