Depth Map Filtering with Adaptive Flying Pixel Thresholds

The patent applies dynamics by transitioning from fixed thresholds to adaptive, dynamically calculated thresholds. The flying pixel determination threshold is no longer a static value but is dynamically adjusted based on local depth map characteristics such as depth gradients and confidence levels at each pixel location. This allows the filtering process to adapt to varying depth discontinuities across different regions of the image, improving accuracy while maintaining computational efficiency through localized adaptive processing.

The patent implements local quality by applying different filtering criteria and threshold values to different regions of the depth map based on local characteristics. Instead of using a uniform threshold across the entire image, the system calculates local depth gradients and confidence levels to determine region-specific thresholds. This enables more accurate identification of flying pixels in areas with high depth discontinuity while preserving valid depth information in homogeneous regions.

The patent applies segmentation by dividing the depth map processing into distinct stages: confidence level calculation, depth gradient computation, threshold determination, and flying pixel identification. Each stage processes specific aspects of the depth data independently, allowing for optimized computational resources at each step. The segmentation also enables parallel processing of different image regions, reducing overall computational power consumption while maintaining adaptive threshold accuracy.

The patent utilizes parameter changes by modifying threshold values based on local depth map parameters such as depth gradients and confidence levels. Instead of using fixed thresholds, the system dynamically adjusts threshold parameters according to local conditions, achieving high accuracy with efficient computation through parameter-driven adaptive filtering rather than complex algorithms.

The patent applies preliminary action by pre-calculating confidence levels and depth gradients for all pixels before performing flying pixel identification. These preliminary computations create ready-to-use parameters that accelerate the main filtering process. By preparing depth gradient maps and confidence level maps in advance, the system reduces the computational burden during the actual flying pixel removal stage, achieving high precision filtering with reduced processing time.

The patent implements skipping by using pre-computed depth gradients and confidence levels to rapidly identify flying pixels without performing complex iterative analysis. The method directly compares local depth differences against adaptive thresholds derived from preliminary calculations, allowing the system to quickly process and filter flying pixels across the entire depth map while maintaining high precision through gradient-based detection.