Zig-Zag Based Single-Pass Connected Components Analysis

Area of Science:

• Computer Vision
• Image Processing
• Algorithm Design

Background:

• Connected Components Analysis (CCA) typically uses raster scans, incurring time overhead for label resolution at row ends.
• Existing CCA algorithms face limitations in processing speed and hardware efficiency.

Purpose of the Study:

• To eliminate the end-of-row label resolution overhead in single-pass CCA.
• To improve the throughput and reduce hardware requirements of CCA algorithms.
• To enable earlier output of feature vectors for completed connected components.

Main Methods:

• Replaced the conventional raster scan with a zig-zag scan for CCA.
• Developed a novel algorithm for label chain resolution during zig-zag scanning.
• Introduced a new method for detecting object completion to minimize latency.

Main Results:

• Eliminated the time overhead associated with end-of-row label resolution.
• Achieved faster processing in worst-case scenarios without end-of-row delays.
• Demonstrated higher image processing throughput compared to state-of-the-art methods.
• Reduced hardware requirements for CCA architectures.

Conclusions:

• The zig-zag scan offers a more efficient approach to single-pass CCA.
• This method significantly improves processing speed and hardware efficiency in image analysis.
• The proposed CCA algorithm presents a viable alternative for high-throughput image processing applications.