Real-time depth-of-field rendering using anisotropically filtered mipmap interpolation

Area of Science:

• Computer Graphics
• Virtual Reality
• Image Processing

Background:

• Real-time rendering of depth-of-field (DoF) effects is crucial for immersive virtual reality (VR) experiences.
• Traditional post-filtering methods often suffer from artifacts like blurring discontinuity and intensity leakage.
• Existing GPU-based techniques require further optimization for artifact reduction and performance scalability.

Purpose of the Study:

• To develop a real-time, GPU-based post-filtering method for rendering high-quality DoF effects in VR.
• To address and eliminate major artifacts commonly found in post-filtering DoF techniques.
• To achieve constant and scalable real-time performance suitable for VR applications.

Main Methods:

• Nonlinear interpolation of mipmap images generated from a pinhole image to achieve blurring.
• Utilizing magnification with a circular filter to mitigate bilinear magnification artifacts.
• Employing anisotropic mipmapping and smoothing of blurring degrees to eliminate intensity leakage and blurring discontinuity.
• Implementing the entire framework using GPU programs for accelerated, real-time performance.

Main Results:

• The proposed method effectively eliminates common post-filtering artifacts, including bilinear magnification artifact, intensity leakage, and blurring discontinuity.
• Achieved constant and scalable real-time rendering performance, essential for virtual reality.
• Demonstrated comparable or superior image quality and rendering performance against recent GPU-based methods.

Conclusions:

• The developed GPU-based post-filtering method provides an effective solution for rendering high-fidelity DoF effects in real-time VR.
• The technique successfully overcomes limitations of previous methods by artifact elimination and performance optimization.
• This approach offers a viable and efficient method for enhancing visual realism in virtual reality environments.