Roughness preserving filter design to remove spatial noise from stereoscopic skin images for stable haptic rendering

Area of Science:

• Computer graphics
• Haptic technology
• Image processing

Background:

• Accurate 3D skin surface reconstruction from stereo images is vital for realistic haptic feedback in touch interactions.
• Spatial noise in disparity maps derived from stereo skin images can cause significant errors in haptic systems.
• Existing noise removal methods often compromise skin roughness, leading to inaccurate texture cloning.

Purpose of the Study:

• To develop a novel noise removal algorithm that preserves skin roughness during 3D surface reconstruction.
• To ensure stable and accurate haptic rendering by minimizing noise without losing texture details.

Main Methods:

• A roughness preserving filter (RPF) algorithm was developed, utilizing singular value decomposition of disparity maps.
• The algorithm employs disparity control (λ) and noise control (k) parameters.
• The optimal noise control parameter (k) is automatically determined based on roughness gradient angles (R_a).

Main Results:

• The RPF algorithm successfully removed spatial noise while preserving skin roughness in real skin images.
• Evaluation using MSE, PSNR, R_a, and R_q demonstrated superior performance compared to a median filter.
• The results indicate the RPF algorithm's potential for stable haptic rendering of skin roughness.

Conclusions:

• The proposed RPF algorithm is a promising solution for filtering stereo skin images without compromising original roughness.
• The algorithm operates automatically and prioritizes roughness parameters (R_a, R_q).
• The RPF method is adaptable to various disparity map building techniques, enhancing its practical applicability.