Method of depth simulation imaging and depth image super-resolution reconstruction for a 2D/3D compatible CMOS image...

Area of Science:

• Computer Vision
• Image Processing
• Sensor Technology

Background:

• Current CMOS image sensors require advanced methods for accurate depth perception.
• Depth imaging parameters significantly influence the quality of captured depth information.
• Super-resolution (SR) techniques are crucial for enhancing the detail in low-resolution depth maps.

Purpose of the Study:

• To develop and validate a depth simulation imaging and depth image super-resolution (SR) method for 2D/3D compatible CMOS image sensors.
• To establish a depth perception model for analyzing imaging parameters and evaluating real-world imaging effects.
• To propose an SR reconstruction algorithm for recovering high-resolution depth maps from low-resolution data.

Main Methods:

• Development of a depth perception model to analyze imaging parameters and effects.
• Verification of the model's validity through depth error analysis, imaging simulation, and physical verification.
• Implementation of a depth SR reconstruction algorithm using depth simulation images on Middlebury and RGB-D datasets.

Main Results:

• The proposed depth SR method achieves optimized recovery effects for depth maps.
• Root Mean Square Error (RMSE) on the Middlebury dataset was as low as 0.0156 m.
• RMSE on the RGB-D dataset was recorded at 0.0223 m.
• The algorithm demonstrated significant RMSE reduction compared to conventional methods, exceeding 16% on Middlebury and 9% on RGB-D datasets.

Conclusions:

• The developed depth simulation and SR method effectively enhances depth map resolution and accuracy for CMOS image sensors.
• The proposed algorithm offers a substantial improvement over existing depth SR techniques.
• This work provides a valuable contribution to advancing depth imaging capabilities in consumer electronics and robotics.