dcv.multiview.stereo.matching

Members

Classes

StereoMatcher

class StereoMatcher

Handles boilerplate code common to all stereo matching methods.

StereoPipeline

class StereoPipeline

This class provides a framework for constructing stereo matching pipelines that conform to the taxonomy laid out in Scharstein and Szeliski (2002).

Functions

absoluteDifference

StereoCostFunction absoluteDifference()

Creates a StereoCostFunction that computes the pixelwise absolute difference between intensities in the left and right images

bilateralDisparityFilter

DisparityRefiner bilateralDisparityFilter(uint windowSize, float sigmaCol, float sigmaSpace)

Applies a bilateral filter to the disparity map in order to correct outliers.

emptyDisparityMap

DisparityMap emptyDisparityMap()

Creates an empty disparity map

medianDisparityFilter

DisparityRefiner medianDisparityFilter(size_t windowSize = 3)

Applies a median filter to the disparity map in order to correct outliers.

normalisedCrossCorrelation

StereoCostFunction normalisedCrossCorrelation(uint windowSize = 5)

Computes the normalised cross correlation, also known as the cosine similarity, between image patches.

semiGlobalAggregator

StereoCostAggregator semiGlobalAggregator(size_t numPaths = 8, CostType p1 = 15, CostType p2 = 100)

Implements the cost aggregation method described by Hirschmuller (2007), commonly known as Semi-Global Matching.

semiGlobalMatchingPipeline

StereoPipeline semiGlobalMatchingPipeline(const ref StereoPipelineProperties props, size_t numPaths = 8, CostType p1 = 15, CostType p2 = 100)

Creates a StereoPipeline that performs semi-global matching.

squaredDifference

StereoCostFunction squaredDifference()

Creates a StereoCostFunction that computes the pixelwise squared difference between intensities in the left and right images

sumAbsoluteDifferences

StereoCostFunction sumAbsoluteDifferences(uint windowSize = 5)

Computes the sum of absolute differences between two image patches in order to compute the matching cost.

winnerTakesAll

DisparityMethod winnerTakesAll()

Implements the naive winner takes all algorithm for computing a diparity map from a cost volume.

Structs

StereoPipelineProperties

struct StereoPipelineProperties

Contains the properties required to build a StereoPipeline

• matching