Gruen.cpp

Go to the documentation of this file.

#include "Chip.h"
#include "GruenTypes.h"
#include "Gruen.h"
#include "DbProfile.h"
#include "Pvl.h"

#include "tnt/tnt_array1d_utils.h"

#include <cmath>
#include <iostream>
#include <sstream>
#include <iomanip>

#include <QTime>


using namespace std;

namespace Isis {

  Gruen::Gruen() : AutoReg(Gruen::getDefaultParameters()) {
    init(Gruen::getDefaultParameters());
  }

  Gruen::Gruen(Pvl &pvl) : AutoReg(pvl) {
    init(pvl);
  }

  void Gruen::WriteSubsearchChips(const QString &pattern) {
    m_filePattern = pattern;
    return;
  }

  void Gruen::setAffineRadio(const AffineRadio &affrad) {
    m_affine = affrad;
    return;
  }

  void Gruen::setAffineRadio() {
    m_affine = getDefaultAffineRadio();
  }

  int Gruen::algorithm(Chip &pattern, Chip &subsearch,
                       const Radiometric &radio, BigInt &ptsUsed,
                       double &resid, GMatrix &atai, AffineRadio &affrad) {

    m_totalIterations++;  //  Bump iteration counter

    // Initialize loop variables
    int tackSamp = pattern.TackSample();
    int tackLine = pattern.TackLine();

    //  Internal variables
    double rshift = radio.Shift();
    double rgain  = radio.Gain();

    int maxPnts((pattern.Samples()-2) * (pattern.Lines()-2));
    GMatrix a(maxPnts, 8);
    GVector lvec(maxPnts);

    //  pattern chip is rh image , subsearch chip is lh image
    resid = 0.0;
    int npts = 0;
    for(int line = 2 ; line < pattern.Lines() ; line++) {
      for(int samp = 2; samp < pattern.Samples() ; samp++) {

        if(!pattern.IsValid(samp, line)) continue;
        if(!subsearch.IsValid(samp, line)) continue;
        if(!subsearch.IsValid(samp + 1, line)) continue;
        if(!subsearch.IsValid(samp - 1, line)) continue;
        if(!subsearch.IsValid(samp, line - 1)) continue;
        if(!subsearch.IsValid(samp, line + 1)) continue;

        //  Sample/Line numbers
        double x0 = (double)(samp - tackSamp);
        double y0 = (double)(line - tackLine);

        // Discrete derivatives (delta sample #)
        double gxtemp = subsearch.GetValue(samp + 1, line) -
                        subsearch.GetValue(samp - 1, line);
        double gytemp = subsearch.GetValue(samp, line + 1) -
                        subsearch.GetValue(samp, line - 1);

        a[npts][0] = gxtemp;
        a[npts][1] = gxtemp * x0;
        a[npts][2] = gxtemp * y0;
        a[npts][3] = gytemp;
        a[npts][4] = gytemp * x0;
        a[npts][5] = gytemp * y0;
        a[npts][6] = 1.0;
        a[npts][7] = subsearch.GetValue(samp, line);

        double ell = pattern.GetValue(samp, line) -
                     (((1.0 + rgain) * subsearch.GetValue(samp, line)) +
                      rshift);

        lvec[npts] = ell;
        resid += (ell * ell);
        npts++;
      }
    }

    // Check for enough points
    ptsUsed = npts;
    if(!ValidPoints(maxPnts, npts)) {
      std::ostringstream mess;
      mess << "Minimum points (" << MinValidPoints(maxPnts)
           << ") criteria not met (" << npts << ")";
      return (logError(NotEnoughPoints, mess.str().c_str()));
    }

    // Create the ATA array
    GMatrix ata(8,8);
    for(int i = 0 ; i < 8 ; i++) {
      for(int j = 0 ; j < 8 ; j++) {
        double temp(0.0);
        for (int k = 0 ; k < npts ; k++) {
          temp += (a[k][i] * a[k][j]);
        }
        ata[i][j] = temp;
      }
    }

    // Solve the ATAI with Cholesky
    try {
      GVector p(8);
      atai = Choldc(ata, p);
      GMatrix b = Identity(8);
      GMatrix x = b;
      atai = Cholsl(atai, p, b, x);
    }
    catch(IException &ie) {
      QString mess = "Cholesky Failed:: " + ie.toString();
      return (logError(CholeskyFailed, mess));
    }

    // Compute the affine update if result are requested by caller.
    GVector atl(8);
    for (int i = 0 ; i < 8 ; i++) {
      double temp(0.0);
      for (int k = 0 ; k < npts ; k++) {
        temp += a[k][i] * lvec[k];
      }
      atl[i] = temp;
    }

    GVector alpha(8, 0.0);
    for (int i = 0 ; i < 8 ; i++) {
      for (int k = 0 ; k < 8 ; k++) {
        alpha[i] += atai[i][k] * atl[k];
      }
    }

   try {
     affrad = AffineRadio(alpha);
    }
    catch(IException &ie) {
      QString mess = "Affine failed: " + ie.toString();
      return (logError(AffineNotInvertable, mess));
    }

    return (0);
  }

  Analysis Gruen::errorAnalysis(const BigInt &npts, const double &resid,
                                const GMatrix &atai) {

    Analysis results;
    results.m_npts = npts;

    // Converged, compute error anaylsis
    double variance = resid / DegreesOfFreedom(npts);
    GMatrix kmat(8,8);
    for(int r = 0 ; r < 8 ; r++) {
      for(int c = 0 ; c < 8 ; c++) {
        kmat[r][c] = variance * atai[r][c];
      }
    }
    results.m_variance = variance;


    // Set up submatrix
    GMatrix skmat(2,2);
    skmat[0][0] = kmat[0][0];
    skmat[0][1] = kmat[0][3];
    skmat[1][0] = kmat[3][0];
    skmat[1][1] = kmat[3][3];
    try {
      GVector eigen(2);
      Jacobi(skmat, eigen, skmat);
      EigenSort(eigen, skmat);
      for (int i = 0 ; i < 2 ; i++) {
        results.m_sevals[i] = eigen[i];
        results.m_kmat[i] = kmat[i*3][i*3];
      }
    }
    catch(IException &ie) {
      QString errmsg = "Eigen Solution Failed:: " + ie.toString();
      results.m_status = logError(EigenSolutionFailed, errmsg);
      return (results);
    }

    results.m_status = 0;
    return (results);
  }

  GMatrix Gruen::Choldc(const GMatrix &a, GVector &p) const {
    int nrows(a.dim1());
    int ncols(a.dim2());

    GMatrix aa = a.copy();
    p = GVector(ncols);

    for(int i = 0 ; i < nrows ; i++) {
      for(int j = i ; j < ncols ; j++) {
        double sum = aa[i][j];
        for(int k = i-1 ; k >= 0 ; k--) {
          sum -= (aa[i][k] * aa[j][k]);
        }
        // Handle diagonal special
        if (i == j) {
          if (sum <= 0.0) {
            throw IException(IException::Programmer,
                             "Choldc failed - matrix not postive definite",
                             _FILEINFO_);
          }
          p[i] = sqrt(sum);
        }
        else {
          aa[j][i] = sum / p[i];
        }
      }
    }
    return (aa);
  }

  GMatrix Gruen::Cholsl(const GMatrix &a,const GVector &p, const GMatrix &b,
                        const GMatrix &x) const {
    assert(b.dim1() == x.dim1());
    assert(b.dim2() == x.dim2());
    assert(p.dim1() == b.dim2());

    int nrows(a.dim1());
    int ncols(a.dim2());

    GMatrix xout = x.copy();
    for(int j = 0 ; j < nrows ; j++) {

      for(int i = 0 ; i < ncols ; i++) {
        double sum = b[j][i];
        for(int k = i-1 ; k >= 0 ; k--) {
          sum -= (a[i][k] * xout[j][k]);
        }
        xout[j][i] = sum / p[i];
      }

      for (int i = ncols-1 ; i >= 0 ; i--) {
        double sum = xout[j][i];
        for(int k = i+1 ; k < ncols ; k++) {
          sum -= (a[k][i] * xout[j][k]);
        }
        xout[j][i] = sum / p[i];
      }

    }
    return (xout);
  }

  int Gruen::Jacobi(const GMatrix &a, GVector &evals,
                    GMatrix &evecs, const int &MaxIters) const {

    int nrows(a.dim1());
    int ncols(a.dim2());
    GMatrix v = Identity(nrows);
    GVector d(nrows),b(nrows), z(nrows);

    for(int ip = 0 ; ip < nrows ; ip++) {
      b[ip] = a[ip][ip];
      d[ip] = b[ip];
      z[ip] = 0.0;
    }

    double n2(double(nrows) * double(nrows));
    GMatrix aa = a.copy();
    int nrot(0);
    for ( ; nrot < MaxIters ; nrot++) {
      double sm(0.0);
      for(int ip = 0 ; ip < nrows-1 ; ip++) {
        for(int iq = ip+1 ; iq < nrows ; iq++) {
          sm += fabs(aa[ip][iq]);
        }
      }

      //  Test for termination condition
      if (sm == 0.0) {
        evals = d;
        evecs = v;
        return (nrot);
      }

      double thresh = (nrot < 3) ? (0.2 * sm/n2 ): 0.0;
      for (int ip = 0 ; ip < nrows-1 ; ip++) {
        for (int iq = ip+1 ; iq < nrows ; iq++) {
          double g = 100.0 * fabs(aa[ip][iq]);
          if ( (nrot > 3) &&
               ( (fabs(d[ip]+g) == fabs(d[ip])) ) &&
               ( (fabs(d[iq]+g) == fabs(d[iq])) ) ) {
              aa[ip][iq] = 0.0;
          }
          else if ( fabs(aa[ip][iq]) > thresh ) {
            double h = d[iq] - d[ip];
            double t;
            if ( (fabs(h)+g) == fabs(h) ) {
              t = aa[ip][iq]/h;
            }
            else {
              double theta = 0.5 * h / aa[ip][iq];
              t = 1.0 / (fabs(theta) + sqrt(1.0 + theta * theta));
              if (theta < 0.0)  t = -1.0 * t;
            }
            double c = 1./sqrt(1.0 + t * t);
            double s = t * c;
            double tau = s / (1.0 + c);

            h = t * aa[ip][iq];
            z[ip] = z[ip] - h;
            z[iq] = z[iq] + h;
            d[ip] = d[ip] - h;
            d[iq] = d[iq] + h;
            aa[ip][iq] = 0.0;

            double g;
            for (int j = 0 ; j < ip-1 ; j++) {
              g = aa[j][ip];
              h = aa[j][iq];
              aa[j][ip] = g - s * (h + g * tau);
              aa[j][iq] = h + s * (g - h * tau);
            }

            for (int j = ip+1 ; j < iq-1 ; j++) {
              g = aa[ip][j];
              h = aa[j][iq];
              aa[ip][j] = g - s * (h + g * tau);
              aa[j][iq] = h + s * (g - h * tau);
            }

            for (int j = iq+1 ; j < ncols ; j++) {
              g = aa[ip][j];
              h = aa[j][iq];
              aa[ip][j] = g - s * (h + g * tau);
              aa[j][iq] = h + s * (g - h * tau);
            }

            for (int j = 0 ; j < ncols ; j++) {
              g = v[j][ip];
              h = v[j][iq];
              v[j][ip] = g - s * (h + g * tau);
              v[j][iq] = h + s * (g - h * tau);
            }
            nrot++;
          }
        }
      }

      for (int ip = 0 ; ip < nrows ; ip++) {
        b[ip] = b[ip] + z[ip];
        d[ip] = b[ip];
        z[ip] = 0.0;
      }
    }

    // Reach here and we have too many iterations
    evals = d;
    evecs = v;
    throw IException(IException::Programmer,
                     "Too many iterations in Jacobi",
                     _FILEINFO_);
    return (nrot);
  }


  GMatrix Gruen::Identity(const int &ndiag) const {
    GMatrix ident(ndiag, ndiag, 0.0);
    for (int i = 0 ; i < ndiag ; i++) {
      ident[i][i] = 1.0;
    }
    return (ident);
  }

  void Gruen::EigenSort(GVector &evals, GMatrix &evecs) const {
    assert(evals.dim1() == evecs.dim1());
    int n(evals.dim1());
    for (int i = 0 ; i < n-1 ; i++ ) {
      int k = i;
      double p = evals[i];
      for (int j = i+1 ; j < n ; j++) {
        if (evals[j] >= p) {
          k = j;
          p = evals[j];
        }
      }
      if (k != i) {
        evals[k] = evals[i];
        evals[i] = p;
        for (int j = 0 ; j < n ; j++) {
          p = evecs[j][i];
          evecs[j][i] = evecs[j][k];
          evecs[j][k] = p;
        }
      }
    }
    return;
  }


  double Gruen::MatchAlgorithm(Chip &pattern, Chip &subsearch) {
    // reset();

    BigInt npts;
    double resid;
    GMatrix atai;
    AffineRadio affrad;
    int status = algorithm(pattern, subsearch, getDefaultRadio(),
                           npts, resid, atai, affrad);
    if (status == 0) {
      //  Compute fit quality
      Analysis analysis = errorAnalysis(npts, resid, atai);
      if (analysis.isValid()) {
        return (analysis.getEigen());
      }
    }

    // Error conditions return failure
    return (Null);
  }

  bool Gruen::CompareFits(double fit1, double fit2) {
    return (fit1 <= fit2);
  }

  AutoReg::RegisterStatus Gruen::Registration(Chip &sChip, Chip &pChip,
      Chip &fChip, int startSamp,
      int startLine, int endSamp,
      int endLine, int bestSamp,
      int bestLine) {
    // Set conditions for writing subsearch chip states per call.  This code
    // implementation ensures caller must request it per call to this routine.
    //  See the WriteSubsearchChips() method.
    m_callCount++;
    QString chipOut = m_filePattern;
    m_filePattern = "";

    //  Initialize match point.  Ensure points are centered to get real cube
    //  line/sample positions
    pChip.SetChipPosition(pChip.TackSample(), pChip.TackLine());
    sChip.SetChipPosition(sChip.TackSample(), sChip.TackLine());
    MatchPoint matchpt = MatchPoint(PointPair(Coordinate(pChip),
                                              Coordinate(sChip)));

    // Create the fit chip whose size is the same as the pattern chip.
    // This chip will contain the final image at the last iteration.  This
    // usage differs from the non-adaptive purpose.
    // It is critical to use the original search chip to create the subsearch.
    // Copying the original search chip and then resizing preserves established
    // minimum/maximum value ranges.  Then, establish chip convergence
    // condition for Gruen's affine.
    fChip = sChip;
    fChip.SetSize(pChip.Samples(), pChip.Lines());
    Threshold thresh(fChip, getAffineTolerance());

    // Set up Affine transform by establishing search tack point
    AffineRadio affine = m_affine;
    m_affine = AffineRadio();  // Reset initial condition for next call

    //  Set up bestLine/bestSample position.  Do this using the local affine
    // and not the search chip.
    Coordinate best(bestLine-sChip.TackLine(), bestSamp-sChip.TackSample());
    affine.Translate(best);


    //  Algorithm parameters
    bool done = false;
    m_nIters = 0;
    do {

      // Extract the sub search chip.  The algorithm method handles the
      // determination of good data volumes
      Affine extractor(affine.m_affine);
      sChip.Extract(fChip, extractor);

      //  If requested for this run, write the current subsearch chip state
      if (!chipOut.isEmpty()) {
        std::ostringstream ss;
        ss << "C" << std::setw(6) << std::setfill('0') << m_callCount << "I"
           << std::setw(3) << std::setfill('0') << m_nIters;
        QString sfname = chipOut + ss.str().c_str() + ".cub";
        fChip.Write(sfname);
      }

      // Try to match the two subchips
      AffineRadio alpha;
      BigInt npts(0);
      GMatrix atai;
      double resid;
      int status = algorithm(pChip, fChip, affine.m_radio, npts, resid,
                             atai, alpha);
      if (status != 0) {
        //  Set failed return condition and give up!
        return (Status(matchpt.setStatus(status)));
      }

      //  Test for termination conditions - errors or convergence
      matchpt.m_nIters = ++m_nIters;
      if (m_nIters > m_maxIters) {
        QString errmsg = "Maximum Iterations exceeded";
        matchpt.setStatus(logError(MaxIterationsExceeded, errmsg));
        return (Status(matchpt));  //  Error condition
      }

      //  Check for convergence after the first pass
      if (m_nIters > 1) {
        if(thresh.hasConverged(alpha)) {
          //  Compute error analysis
          matchpt.m_affine = affine;
          matchpt.m_analysis = errorAnalysis(npts, resid, atai);
          matchpt.setStatus(matchpt.m_analysis.m_status);
          if (matchpt.isValid()) {
            //  Update the point even if constraints don't pass
            Coordinate uCoord = getChipUpdate(sChip, matchpt);
            SetChipSample(uCoord.getSample());
            SetChipLine(uCoord.getLine());
            SetGoodnessOfFit(matchpt.getEigen());

            // Check constraints
            matchpt.setStatus(CheckConstraints(matchpt));
          }

          //  Set output point
          m_point = matchpt;
          return (Status(matchpt));  // with AutoReg status
        }
      }
        //  Not done yet - update the affine transform for next loop
      try {
        affine += alpha;
      }
      catch (IException &ie) {
        QString mess = "Affine invalid/not invertable";
        matchpt.setStatus(logError(AffineNotInvertable, mess));
        return (Status(matchpt));  //  Another error condition to return
      }
    } while (!done);

    return (Status(matchpt));
  }

  Pvl &Gruen::getDefaultParameters() {
    static Pvl regdef;
    regdef = Pvl("$base/templates/autoreg/coreg.adaptgruen.p1515s3030.def");
    return (regdef);
  }

  Pvl Gruen::AlgorithmStatistics(Pvl &pvl) {
    PvlGroup algo("GruenFailures");
    algo += PvlKeyword("Name", AlgorithmName());
    algo += PvlKeyword("Mode", "Adaptive");

    //  Log errors
    for (int e = 0 ; e <  m_errors.size() ; e++) {
      algo += m_errors.getNth(e).LogIt();
    }

    if (m_unclassified > 0) {
      algo += PvlKeyword("UnclassifiedErrors", toString(m_unclassified));
    }
    pvl.addGroup(algo);
    pvl.addGroup(StatsLog());
    pvl.addGroup(ParameterLog());
    return (pvl);
  }

  PvlGroup Gruen::StatsLog() const {
    PvlGroup stats("GruenStatistics");

    stats += PvlKeyword("TotalIterations",   toString(m_totalIterations));
    stats += ValidateKey("IterationMinimum", m_iterStat.Minimum());
    stats += ValidateKey("IterationAverage", m_iterStat.Average());
    stats += ValidateKey("IterationMaximum", m_iterStat.Maximum());
    stats += ValidateKey("IterationStandardDeviation", m_iterStat.StandardDeviation());

    stats += ValidateKey("EigenMinimum", m_eigenStat.Minimum());
    stats += ValidateKey("EigenAverage", m_eigenStat.Average());
    stats += ValidateKey("EigenMaximum", m_eigenStat.Maximum());
    stats += ValidateKey("EigenStandardDeviation", m_eigenStat.StandardDeviation());

    stats += ValidateKey("RadioShiftMinimum", m_shiftStat.Minimum());
    stats += ValidateKey("RadioShiftAverage", m_shiftStat.Average());
    stats += ValidateKey("RadioShiftMaximum", m_shiftStat.Maximum());
    stats += ValidateKey("RadioShiftStandardDeviation", m_shiftStat.StandardDeviation());

    stats += ValidateKey("RadioGainMinimum", m_gainStat.Minimum());
    stats += ValidateKey("RadioGainAverage", m_gainStat.Average());
    stats += ValidateKey("RadioGainMaximum", m_gainStat.Maximum());
    stats += ValidateKey("RadioGainStandardDeviation", m_gainStat.StandardDeviation());

    return (stats);

  }

  PvlGroup Gruen::ParameterLog() const {
    PvlGroup parms("GruenParameters");

    parms += PvlKeyword("MaximumIterations", toString(m_maxIters));
    parms += ValidateKey("AffineScaleTolerance", m_scaleTol);
    parms += ValidateKey("AffineShearTolerance", m_shearTol);
    parms += ValidateKey("AffineTranslationTolerance", m_transTol);

    parms += ParameterKey("AffineTolerance", m_affineTol);
    parms += ParameterKey("SpiceTolerance",  m_spiceTol);

    parms += ParameterKey("RadioShiftTolerance", m_shiftTol);

    parms += ParameterKey("RadioGainMinTolerance", m_rgainMinTol);
    parms += ParameterKey("RadioGainMaxTolerance", m_rgainMaxTol);

    parms += ValidateKey("DefaultRadioGain",  m_defGain);
    parms += ValidateKey("DefaultRadioShift", m_defShift);

    return (parms);
  }


  Gruen::ErrorList Gruen::initErrorList() const {
    ErrorList elist;
    elist.add(1, ErrorCounter(1, "NotEnoughPoints"));
    elist.add(2, ErrorCounter(2, "CholeskyFailed"));
    elist.add(3, ErrorCounter(3, "EigenSolutionFailed"));
    elist.add(4, ErrorCounter(4, "AffineNotInvertable"));
    elist.add(5, ErrorCounter(5, "MaxIterationsExceeded"));
    elist.add(6, ErrorCounter(6, "RadShiftExceeded"));
    elist.add(7, ErrorCounter(7, "RadGainExceeded"));
    elist.add(8, ErrorCounter(8, "MaxEigenExceeded"));
    elist.add(9, ErrorCounter(9, "AffineDistExceeded"));
    return (elist);
  }


  int Gruen::logError(int gerrno, const QString &gerrmsg) {
    if (!m_errors.exists(gerrno)) {
      m_unclassified++;
    }
    else {
      m_errors.get(gerrno).BumpIt();
    }
    return (gerrno);
  }

  void Gruen::init(Pvl &pvl) {
    //  Establish the parameters
    if (pvl.hasObject("AutoRegistration")) {
      m_prof = DbProfile(pvl.findGroup("Algorithm", Pvl::Traverse));
    }
    else {
      m_prof = DbProfile(pvl);
    }

    if (m_prof.Name().isEmpty())  m_prof.setName("Gruen");

    // Define internal parameters
    m_maxIters = toInt(ConfKey(m_prof, "MaximumIterations", toString(30)));

    m_transTol = toDouble(ConfKey(m_prof, "AffineTranslationTolerance", toString(0.1)));
    m_scaleTol = toDouble(ConfKey(m_prof, "AffineScaleTolerance", toString(0.3)));
    m_shearTol = toDouble(ConfKey(m_prof, "AffineShearTolerance", toString(m_scaleTol)));
    m_affineTol = toDouble(ConfKey(m_prof, "AffineTolerance", toString(DBL_MAX)));

    m_spiceTol = toDouble(ConfKey(m_prof, "SpiceTolerance", toString(DBL_MAX)));

    m_shiftTol = toDouble(ConfKey(m_prof, "RadioShiftTolerance", toString(DBL_MAX)));
    m_rgainMinTol = toDouble(ConfKey(m_prof, "RadioGainMinTolerance", toString(-DBL_MAX)));
    m_rgainMaxTol = toDouble(ConfKey(m_prof, "RadioGainMaxTolerance", toString(DBL_MAX)));

    // Set radiometric defaults
    m_defGain  =  toDouble(ConfKey(m_prof, "DefaultRadioGain", toString(0.0)));
    m_defShift =  toDouble(ConfKey(m_prof, "DefaultRadioShift", toString(0.0)));

    m_callCount = 0;
    m_filePattern = "";

    m_nIters = 0;
    m_totalIterations = 0;

    m_errors = initErrorList();
    m_unclassified = 0;

    m_defAffine = AffineRadio(getDefaultRadio());
    m_affine = getAffineRadio();
    m_point  = MatchPoint(m_affine);

    //reset();
    return;
  }


  void Gruen::resetStats() {
    m_eigenStat.Reset();
    m_iterStat.Reset();
    m_shiftStat.Reset();
    m_gainStat.Reset();
    return;
  }

  BigInt Gruen::MinValidPoints(BigInt totalPoints) const {
    double pts = (double) totalPoints * (PatternValidPercent() / 100.0);
    return ((BigInt) pts);
  }

  bool Gruen::ValidPoints(BigInt totalPoints, BigInt nPoints) const {
    return (nPoints > MinValidPoints(totalPoints));
  }

  AffineTolerance Gruen::getAffineTolerance() const {
    return (AffineTolerance(m_transTol, m_scaleTol, m_shearTol));
  }

  int Gruen::CheckConstraints(MatchPoint &point) {

    //  Point must be good for check to occur
    if (point.isValid()) {
      if (point.m_nIters > m_maxIters) {
        QString errmsg = "Maximum Iterations exceeded";
        return (logError(MaxIterationsExceeded, errmsg));
      }
      m_iterStat.AddData(point.m_nIters);

      if (point.getEigen() > Tolerance()) {
        QString errmsg = "Maximum Eigenvalue exceeded";
        return (logError(MaxEigenExceeded, errmsg));
      }
      m_eigenStat.AddData(point.getEigen());

      double shift = point.m_affine.m_radio.Shift();
      if ( shift > m_shiftTol) {
        QString errmsg = "Radiometric shift exceeds tolerance";
        return (logError(RadShiftExceeded, errmsg));
      }
      m_shiftStat.AddData(shift);

      double gain = point.m_affine.m_radio.Gain();
      if (((1.0 + gain) > m_rgainMaxTol) ||
          ((1.0 + gain) < m_rgainMinTol)) {
        QString errmsg = "Radiometric gain exceeds tolerances";
        return (logError(RadGainExceeded, errmsg));
      }
      m_gainStat.AddData(gain);


      double dist = point.getAffinePoint(Coordinate(0.0, 0.0)).getDistance();
      if (dist > getAffineConstraint()) {
        QString errmsg = "Affine distance exceeded";
        return (logError(AffineDistExceeded, errmsg));
      }
    }
    return (point.getStatus());
  }

  Coordinate Gruen::getChipUpdate(Chip &chip, MatchPoint &point) const {
    Coordinate chippt = point.getAffinePoint(Coordinate(0.0, 0.0));
    chip.SetChipPosition(chip.TackSample(), chip.TackLine());
    chip.TackCube(chip.CubeSample()+chippt.getSample(),
                  chip.CubeLine()+chippt.getLine());
    return (Coordinate(chip.ChipLine(), chip.ChipSample()));
  }


  AutoReg::RegisterStatus Gruen::Status(const MatchPoint &mpt) {
    if ( mpt.isValid() ) { return (AutoReg::SuccessSubPixel);  }
    return (AutoReg::AdaptiveAlgorithmFailed);
  }

} // namespace Isis