Matched filter processing of raster-scanned images: analysis

Area of Science:

• Image processing
• Signal processing
• Digital imaging

Background:

• Matched filter techniques are crucial for object detection in digital images.
• TV raster-scanned images present unique challenges due to their scanning method.
• Understanding filter performance is key to optimizing image analysis.

Purpose of the Study:

• To analyze and simulate the performance of matched filters on TV raster-scanned images.
• To investigate the impact of filter bandwidth on output correlation.
• To determine the relationship between object size, scan lines, and correlation performance.

Main Methods:

• Simulation of matched filter technique using two input objects.
• Analysis of output correlation as a function of matched filter bandwidth.
• Evaluation of performance under varying scan line densities.

Main Results:

• For Nyquist bandwidth filters, output correlation shows minimal decrease (3-4%) with smaller objects.
• Substantial correlation decrease occurs when matched filter bandwidth covers the total object spectrum and scan lines per element are below four.
• Bandwidth selection is critical for maintaining correlation fidelity.

Conclusions:

• Matched filter performance is robust for smaller objects within Nyquist bandwidth limits.
• Exceeding object spectrum bandwidth and low scan line density significantly degrade correlation.
• Optimizing bandwidth is essential for accurate object detection in raster-scanned images.