Related Experiment Videos
Stationary and nonstationary spatial domain Metz filtering
M A King1, T R Miller, P W Doherty
1Department of Nuclear Medicine, University of Massachusetts Medical School, Worcester 01605.
Nuclear Medicine Communications
|January 1, 1988
Summary
This study explored stationary and nonstationary finite-impulse-response (FIR) Metz filters for image processing. While both improved tumor detection, nonstationary FIR filtering offered no significant advantage over stationary methods.
Area of Science:
- Medical image analysis
- Digital signal processing
Background:
- The count-dependent Metz filter is a tool for image processing.
- Finite-impulse-response (FIR) implementations can be stationary or nonstationary.
Purpose of the Study:
- To investigate stationary and nonstationary FIR implementations of the count-dependent Metz filter.
- To assess the impact of filter size on image quality.
- To evaluate the effectiveness of these filters in 'tumor' detection.
Main Methods:
- Developed a nonstationary FIR Metz filter using an algorithm to select preformed FIR filters based on pixel count.
- Compared the execution time of stationary and nonstationary FIR filtering.
- Assessed 'tumor' detection rates using unprocessed images, stationary FIR Metz filtering, and nonstationary FIR Metz filtering.
Main Results:
- Filter size is critical for image quality; a 15x15 FIR filter is recommended for 128x128 images.
- Nonstationary FIR filtering required minimal increase in execution time compared to stationary filtering.
- All Metz filtering techniques significantly improved 'tumor' detection compared to unprocessed images.
Conclusions:
- Nonstationary FIR Metz filtering does not offer a significant advantage over stationary FIR Metz filtering for 'tumor' detection when filters are optimized solely on pixel count.
- Filter size is a key determinant of image quality in Metz filtering.