Interpolate2D.h

Classes

Interpolate2D -- A two dimension interpolator for Matrices or Arrays (full description)

class Interpolate2D

Types

enum Method

NEAREST
Nearest neighbour

LINEAR
Bilinear

CUBIC
Bicubic

Interface

Public Members

Interpolate2D(Interpolate2D::Method method=Interpolate2D::LINEAR)

Interpolate2D(const Interpolate2D &other)

~Interpolate2D()

Interpolate2D &operator=(const Interpolate2D &other)

Bool interp (Float &result, const Vector<Double> &where, const Matrix<Float> &data) const

Bool interp (Float &result, const Vector<Double> &where, const Matrix<Float> &data, const Matrix<Bool> &mask) const

Bool interp (Double &result, const Vector<Double> &where, const Matrix<Double> &data) const

Bool interp (Double &result, const Vector<Double> &where, const Matrix<Double> &data, const Matrix<Bool> &mask) const

Bool interp(Double &resultI, Double &resultJ, const Vector<Double> &where, const Matrix<Double> &dataI, const Matrix<Double> &dataJ, const Matrix<Bool> &mask) const

template <typename T> Bool interpLinear2(T &resultI, T &resultJ, const T<Double> &where, const Vector<T> &dataI, const Vector<T> &dataJ, const Vector<Bool> &mask) const

Bool interp (Bool &result, const Vector<Double> &where, const Matrix<Bool> &data) const

Method interpolationMethod() const

static Interpolate2D::Method stringToMethod(const String &method)

Private Members

Bool anyBadMaskPixels (const Matrix<Bool>* &mask, Int i1, Int i2, Int j1, Int j2) const

template <typename T> Bool interpNearest(T &result, const T<Double> &where, const Vector<T> &data, const Vector<Bool>* &maskPtr) const

Bool interpNearestBool(Bool &result, const Vector<Double> &where, const Matrix<Bool> &data) const

template <typename T> Bool interpLinear(T &result, const T<Double> &where, const Vector<T> &data, const Vector<Bool>* &maskPtr) const

Bool interpLinearBool(Bool &result, const Vector<Double> &where, const Matrix<Bool> &data) const

template <typename T> Bool interpCubic(T &result, const T<Double> &where, const Vector<T> &data, const Vector<Bool>* &maskPtr) const

Bool interpCubicBool(Bool &result, const Vector<Double> &where, const Matrix<Bool> &data) const

void bcucof (Double c[4][4], const Double y[4], const Double y1[4], const Double y2[4], const Double y12[4]) const

typedef Bool(Interpolate2D::*FuncPtrFloat) (Float &result, const FuncPtr<Double> &where, const Vector<Float> &data, const Vector<Bool>* &maskPtr) const

typedef Bool(Interpolate2D::*FuncPtrDouble) (Double &result, const FuncPtr<Double> &where, const Vector<Double> &data, const Vector<Bool>* &maskPtr) const

typedef Bool(Interpolate2D::*FuncPtrBool) (Bool &result, const FuncPtr<Double> &where, const Vector<Bool> &data) const

Description

Review Status

Reviewed By:: wbrouw
Date Reviewed:: 2004/05/26

Prerequisite

Arrays

Etymology

This class is called Interpolate2D because it does 2 dimensional interpolations

Synopsis

Given a regular Array or Matrix and a vector of pixel coordinates, interpolate the values of that array/matrix onto those pixel coordinates.

Absolutely no checking of the consistency of the input data is done in order to preserve maximum speed. The coordinate vector *must* have at least 2 elements (others will be ignored). If you supply data and mask, those arrays *must* be the same shape. Failure to follow these rules will result in your program crashing.

Example

    
    Matrix<Float> matt(10,10);
    Vector<Float> where(2);  
    where(0) = 3.452;  where(1) = 6.1;
    Interpolate2D myInterp(Interpolate2D::LINEAR);
    Float result;
    Bool ok = myInterp(result, where, matt);

Motivation

2-D interpolation is required in geometry transformation routines such as in ImageRegrid.

To Do

Now that there are float/double/bool versions, the class should be templated and specialized versions made as needed. The code duplucation in the Float/Double versions is pretty awful presently.
Alternative approach: instantiate with an Array, take a block of vector locations, return a block of interpolation results

Member Description

enum Method

Interpolate2D(Interpolate2D::Method method=Interpolate2D::LINEAR)

Constructor

Interpolate2D(const Interpolate2D &other)

Copy constructor (copy semantics)

~Interpolate2D()

destructor

Interpolate2D &operator=(const Interpolate2D &other)

Assignment operator (copy semantics)

Bool interp (Float &result, const Vector<Double> &where, const Matrix<Float> &data) const
Bool interp (Float &result, const Vector<Double> &where, const Matrix<Float> &data, const Matrix<Bool> &mask) const

Do one Float interpolation, supply Matrix and mask (True is good), and pixel coordinate. Returns False if coordinate out of range or data are masked. No shape integrity checking is done (see above).

Bool interp (Double &result, const Vector<Double> &where, const Matrix<Double> &data) const
Bool interp (Double &result, const Vector<Double> &where, const Matrix<Double> &data, const Matrix<Bool> &mask) const

Do one Double interpolation, supply Matrix/Array and mask (True is good), and pixel coordinate. Returns False if coordinate out of range or data are masked. No shape integrity checking is done (see above).

Bool interp(Double &resultI, Double &resultJ, const Vector<Double> &where, const Matrix<Double> &dataI, const Matrix<Double> &dataJ, const Matrix<Bool> &mask) const
template <typename T> Bool interpLinear2(T &resultI, T &resultJ, const T<Double> &where, const Vector<T> &dataI, const Vector<T> &dataJ, const Vector<Bool> &mask) const

Do two linear interpolations simultaneously. The second call is direct. The first call transfers to the second call. It is assumed that the structure (shape, steps) of the mask and data files are the same.

Bool interp (Bool &result, const Vector<Double> &where, const Matrix<Bool> &data) const

Do one interpolation, supply boolean Matrix (True is good), and pixel coordinate. Returns False if coordinate out of range. The result is False if any data value in the interpolation grid are False (bad), else True. No shape integrity checking is done.

Method interpolationMethod() const

Recover interpolation method

static Interpolate2D::Method stringToMethod(const String &method)

Convert string ("nearest", "linear", "cubic") to interpolation method Minimum match will do.

Bool anyBadMaskPixels (const Matrix<Bool>* &mask, Int i1, Int i2, Int j1, Int j2) const

Are any of the mask pixels bad ? Returns False if no mask.

template <typename T> Bool interpNearest(T &result, const T<Double> &where, const Vector<T> &data, const Vector<Bool>* &maskPtr) const

nearest neighbour interpolation

Bool interpNearestBool(Bool &result, const Vector<Double> &where, const Matrix<Bool> &data) const

template <typename T> Bool interpLinear(T &result, const T<Double> &where, const Vector<T> &data, const Vector<Bool>* &maskPtr) const

bi-linear interpolation

Bool interpLinearBool(Bool &result, const Vector<Double> &where, const Matrix<Bool> &data) const

template <typename T> Bool interpCubic(T &result, const T<Double> &where, const Vector<T> &data, const Vector<Bool>* &maskPtr) const

bi-cubic interpolation

Bool interpCubicBool(Bool &result, const Vector<Double> &where, const Matrix<Bool> &data) const

void bcucof (Double c[4][4], const Double y[4], const Double y1[4], const Double y2[4], const Double y12[4]) const

helping routine from numerical recipes

typedef Bool(Interpolate2D::FuncPtrFloat) (Float &result, const FuncPtr<Double> &where, const Vector<Float> &data, const Vector<Bool> &maskPtr) const

Typedefs for function pointers

typedef Bool(Interpolate2D::FuncPtrDouble) (Double &result, const FuncPtr<Double> &where, const Vector<Double> &data, const Vector<Bool> &maskPtr) const