Class ConfigDisparityBM

All Implemented Interfaces:
Configuration, Serializable
Direct Known Subclasses:

public class ConfigDisparityBM extends Object implements Configuration
Configuration for the basic block matching stereo algorithm that employs a greedy winner takes all strategy.
See Also:
  • Field Details

    • disparityMin

      public int disparityMin
      Minimum disparity that it will check. Must be ≥ 0 and < maxDisparity
    • disparityRange

      public int disparityRange
      Number of disparity values considered. Must be > 0
    • regionRadiusX

      public int regionRadiusX
      Radius of the rectangular region along x-axis.
    • regionRadiusY

      public int regionRadiusY
      Radius of the rectangular region along y-axis.
    • maxPerPixelError

      public double maxPerPixelError
      Maximum allowed error in a region per pixel. Only used by "error" based measures, e.g. NCC does not use this value. Set to < 0 to disable.
    • validateRtoL

      public int validateRtoL
      Tolerance for how difference the left to right associated values can be. Try 1. Disable with -1
    • texture

      public double texture
      Tolerance for how similar optimal region is to other region. Closer to zero is more tolerant. Try 0.15 unless NCC then 0.005. Disable with a value ≤ 0
    • subpixel

      public boolean subpixel
      If subpixel should be used to find disparity or not. If on then output disparity image needs to me GrayF32. If false then GrayU8.
    • saveScore

      public boolean saveScore
      If true it will save the disparity score for each pixel
    • errorType

      public DisparityError errorType
      How the error is computed for each block
    • configCensus

      public ConfigDisparityError.Census configCensus
      Used if error type is Census
    • configNCC

      public ConfigDisparityError.NCC configNCC
      Used if error type is NCC
    • border

      public BorderType border
      Specifies how the image border is handled. In general you want to avoid an approach which would bias the error to prefer a region with lots of pixels outside the image border.
    • catastrophicReset

      public int catastrophicReset
      This recomputes from scratch the disparity score every N rows. If this is not done then large errors can build up resulting in inaccurate range estimates. This is hardly noticeable in most stereo benchmarks as everything is up close. Set to 1 to maximize numerical stability and set to Integer.MAX_VALUE to disable.
      See Also:
  • Constructor Details

    • ConfigDisparityBM

      public ConfigDisparityBM()
  • Method Details