Video and Image Processing Suite User Guide

ID 683416
Date 4/04/2022
Public
Document Table of Contents

13.1.3. Filtered

The filtered algorithm is the most computationally expensive and device resource heavy algorithm, but it offers increased visual quality.

You can parameterize the filtered algorithm to use either left siting (co-siting) or center siting of the chroma data.

  • For downsampling conversions (4:4:4 to 4:2:2), the filtered algorithm applies an 8-tap Lanczos-2 resampling filter to generate the downsampled data. Different phase shifts are applied to the Lanczos-2 function when generating the coefficients, depending on the siting selected and whether the pixel index is even or odd. For left chroma siting, phase shifts of 0 and 0.5 are applied to the Lanczos-2 coefficients for the,even and odd indexed chroma samples respectively. For center chroma siting, the phases shifts are –0.25 and +0.25.
  • For upsampling conversions (4:2:2 to 4:4:4), the filtered algorithm applies a 4-tap Lanczos-2 resampling filter to generate the upsampled data. For left chroma siting phase shifts of 0 and 0.5 are applied to the Lanczos-2 coefficients for the even and odd indexed chroma samples respectively. For center chroma siting the phases shifts are -0.25 and +0.25.

You may also opt to enable luma adaption for upsampling conversions. This feature further increases device resource usage (and is the only chroma resampler mode to implement some logic in DSP blocks), but may reduce color bleed around edges when compared to the default filtered algorithm.

When you enable luma adaption, the differences between successive luma samples are computed and compared to an edge threshold to detect significant edges. In areas where edges with strong vertical components are detected the phase of the Lanczos-2 filter can be shifted by up to 0.25 to the left or right to weight the resulting chroma samples more heavily towards the more appropriate side of the edge.