From theory to application: fast, universal, and cost-effective pixel-level surface reconstruction in shift-rotation...

Area of Science:

• Computer Vision
• Metrology
• Computational Imaging

Background:

• Traditional surface reconstruction methods for shift-rotation absolute measurements face computational challenges due to large matrix computations.
• Existing techniques require significant storage and processing power, limiting their application on common hardware.

Purpose of the Study:

• To develop a more efficient and resource-friendly pixel-level surface reconstruction algorithm.
• To overcome the computational bottlenecks of existing methods in absolute measurements.
• To provide a universally adaptable algorithm for various translation amounts.

Main Methods:

• Frequency domain analysis to derive guidelines for translation amount selection.
• A novel pixel-level surface reconstruction algorithm based on inverse iteration.
• Elimination of large-scale coefficient matrix construction, storage, and computation.

Main Results:

• Reduced time complexity from cubic to linear.
• Reduced space complexity from quadratic to linear.
• Substantially decreased computational resource requirements and time costs.
• Demonstrated superior accuracy and efficiency compared to state-of-the-art methods.

Conclusions:

• The proposed inverse iteration-based algorithm offers a computationally efficient and accurate solution for pixel-level surface reconstruction.
• This method significantly lowers hardware requirements, broadening the applicability of advanced surface reconstruction techniques.
• The algorithm's adaptability to moderate translation amounts enhances its practical utility in absolute measurements.