Jitter noise modeling and its removal using recursive least squares in shape from focus systems

Area of Science:

• Computer Vision
• 3D Reconstruction
• Image Processing

Background:

• Shape from Focus (SFF) is a passive 3D shape recovery technique using a single viewpoint.
• SFF relies on acquiring a stack of images with varying focus settings.
• Mechanical vibrations in practical SFF systems introduce 'jitter noise' into focus curves, degrading 3D shape accuracy.

Purpose of the Study:

• To model and mitigate jitter noise in SFF systems.
• To develop a robust method for accurate 3D shape recovery despite focus sampling errors.
• To introduce a new metric for evaluating 3D reconstruction quality.

Main Methods:

• Formulated a mathematical model for jitter noise using quadratic functions and Taylor series.
• Applied recursive least squares (RLS) filtering to address the jittering problem.
• Introduced a novel 'depth distortion' (DD) metric to quantify reconstruction errors.

Main Results:

• The proposed RLS filtering effectively reduces jitter noise in SFF systems.
• Experiments on real and simulated objects demonstrated significant improvements in 3D shape recovery.
• The new DD metric, along with RMSE and correlation, accurately assessed reconstructed shape quality.

Conclusions:

• The developed method successfully overcomes the limitations of traditional noise removal techniques for SFF.
• The proposed scheme offers a robust solution for accurate 3D shape recovery in the presence of jitter noise.
• The findings confirm the effectiveness of the RLS filtering approach and the utility of the DD metric.