Dynamic Vision Sensor Disparity Acquisition via Event Segmentation

The patent segments the image processing task by dividing pixels into multiple blocks and selecting representative pixels within each block. Instead of processing all pixels individually, the method processes only the selected representative pixels from each block, significantly reducing the total number of pixels to be processed while maintaining disparity calculation accuracy through the representative nature of selected pixels.

The patent applies local quality by differentiating the processing approach between different regions. Representative pixels are selected based on local characteristics within each block, and different matching strategies are applied: exhaustive search for representative pixels and gradient descent for non-representative pixels. This localized differentiation optimizes calculation efficiency while preserving measurement precision.

The patent reduces processing complexity by segmenting the image into blocks and selecting only representative pixels from each block for exhaustive matching. This segmentation approach simplifies the overall processing complexity while maintaining accuracy by focusing computational resources on key representative pixels rather than all pixels uniformly.

The patent applies partial action by performing exhaustive search only for selected representative pixels while using gradient descent for non-representative pixels. This partial application of the more complex exhaustive search method reduces overall device complexity while maintaining measurement precision through the strategic selection of which pixels receive intensive processing.

The patent improves productivity by segmenting pixels into representative and non-representative categories within blocks. Only representative pixels undergo exhaustive search, while non-representative pixels use gradient descent. This segmentation maintains disparity prediction precision for critical pixels while significantly improving overall processing efficiency by reducing the number of pixels requiring intensive exhaustive search.

The patent applies partial exhaustive search action only to representative pixels rather than all pixels. This partial application of exhaustive search maintains disparity prediction precision where it matters most (at representative pixels) while improving productivity by using the faster gradient descent method for remaining pixels, achieving a balance between precision and efficiency.