New Approaches for Monitoring Image Data.

Area of Science:

• Statistical Process Control
• Image Analysis
• High-Dimensional Data

Background:

• Monitoring high-resolution images with spatially correlated pixels presents challenges due to data dimensionality.
• Direct pixel monitoring is computationally infeasible.
• Existing methods often involve computationally intensive and unstable matrix inversions.

Purpose of the Study:

• To develop new, computationally efficient, and stable control charts for monitoring image processes.
• To address the challenges of high-dimensional and spatially correlated data in image monitoring.
• To propose modifications of the generalized likelihood ratio statistic for improved performance.

Main Methods:

• Development of control charts based on regions of interest for dimension reduction.
• Application of residual charts using the generalized likelihood ratio approach.
• Introduction of two modified generalized likelihood ratio statistics for enhanced stability and efficiency.

Main Results:

• The proposed control charts demonstrate improved performance in high-dimensional settings.
• The new methods offer a solution to the computational and stability issues of existing techniques.
• Simulation studies confirm the effectiveness of the developed control charts.

Conclusions:

• The novel control charts provide a viable and efficient approach for monitoring image processes with spatially correlated pixels.
• These techniques offer practical advantages over existing methods in terms of computation time and result stability.
• The study contributes advanced statistical tools for quality control in high-resolution imaging applications.