annotate base/ColumnOp.cpp @ 1394:9ef1cc26024c

Add Range01 normalisation method to ColumnOp. This is the normalisation that is actually used in the Colour 3D Plot layer historically when column normalisation is enabled (not Max1 after all).

/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */

/*
    Sonic Visualiser
    An audio file viewer and annotation editor.
    Centre for Digital Music, Queen Mary, University of London.
    This file copyright 2006-2016 Chris Cannam and QMUL.
    
    This program is free software; you can redistribute it and/or
    modify it under the terms of the GNU General Public License as
    published by the Free Software Foundation; either version 2 of the
    License, or (at your option) any later version.  See the file
    COPYING included with this distribution for more information.
*/

#include "ColumnOp.h"

#include <cmath>
#include <algorithm>
#include <iostream>

#include "base/Debug.h"

using namespace std;

ColumnOp::Column
ColumnOp::fftScale(const Column &in, int fftSize)
{
    return applyGain(in, 2.0 / fftSize);
}

ColumnOp::Column
ColumnOp::peakPick(const Column &in)
{
    vector<float> out(in.size(), 0.f);

    for (int i = 0; in_range_for(in, i); ++i) {
        if (isPeak(in, i)) {
            out[i] = in[i];
        }
    }
    
    return out;
}

ColumnOp::Column
ColumnOp::normalize(const Column &in, ColumnNormalization n) {

    if (n == ColumnNormalization::None || in.empty()) {
        return in;
    }
    
    float shift = 0.f;
    float scale = 1.f;

    if (n == ColumnNormalization::Range01) {

        float min = 0.f;
        float max = 0.f;
        bool have = false;
        for (auto v: in) {
            if (v < min || !have) {
                min = v;
            }
            if (v > max || !have) {
                max = v;
            }
            have = true;
        }
        if (min != 0.f) {
            shift = -min;
            max -= min;
        }
        if (max != 0.f) {
            scale = 1.f / max;
        }

    } else if (n == ColumnNormalization::Sum1) {

        float sum = 0.f;

        for (auto v: in) {
            sum += fabsf(v);
        }

        if (sum != 0.f) {
            scale = 1.f / sum;
        }

    } else {

        float max = 0.f;

        for (auto v: in) {
            v = fabsf(v);
            if (v > max) {
                max = v;
            }
        }

        if (n == ColumnNormalization::Max1) {
            if (max != 0.f) {
                scale = 1.f / max;
            }
        } else if (n == ColumnNormalization::Hybrid) {
            if (max > 0.f) {
                scale = log10f(max + 1.f) / max;
            }
        }
    }

    return applyGain(applyShift(in, shift), scale);
}

ColumnOp::Column
ColumnOp::distribute(const Column &in,
                     int h,
                     const vector<double> &binfory,
                     int minbin,
                     bool interpolate)
{
    vector<float> out(h, 0.f);
    int bins = int(in.size());

    if (interpolate) {
        // If the bins are all closer together than the target y
        // coordinate increments, then we don't want to interpolate
        // after all. But because the binfory mapping isn't
        // necessarily linear, just checking e.g. whether bins > h is
        // not enough -- the bins could still be spaced more widely at
        // either end of the scale. We are prepared to assume however
        // that if the bins are closer at both ends of the scale, they
        // aren't going to diverge mysteriously in the middle.
        if (h > 1 &&
            fabs(binfory[1] - binfory[0]) >= 1.0 &&
            fabs(binfory[h-1] - binfory[h-2]) >= 1.0) {
            interpolate = false;
        }
    }
    
    for (int y = 0; y < h; ++y) {

        if (interpolate) {

            double sy = binfory[y] - minbin - 0.5;
            double syf = floor(sy);

            int mainbin = int(syf);
            int other = mainbin;
            if (sy > syf) {
                other = mainbin + 1;
            } else if (sy < syf) {
                other = mainbin - 1;
            }

            if (mainbin < 0) {
                mainbin = 0;
            }
            if (mainbin >= bins) {
                mainbin = bins - 1;
            }

            if (other < 0) {
                other = 0;
            }
            if (other >= bins) {
                other = bins - 1;
            }

            double prop = 1.0 - fabs(sy - syf);
            
            double v0 = in[mainbin];
            double v1 = in[other];
                
            out[y] = float(prop * v0 + (1.0 - prop) * v1);

        } else {
            
            double sy0 = binfory[y] - minbin;

            double sy1;
            if (y+1 < h) {
                sy1 = binfory[y+1] - minbin;
            } else {
                sy1 = bins;
            }

            int by0 = int(sy0 + 0.0001);
            int by1 = int(sy1 + 0.0001);

            if (by0 < 0 || by0 >= bins || by1 > bins) {
                SVCERR << "ERROR: bin index out of range in ColumnOp::distribute: by0 = " << by0 << ", by1 = " << by1 << ", sy0 = " << sy0 << ", sy1 = " << sy1 << ", y = " << y << ", binfory[y] = " << binfory[y] << ", minbin = " << minbin << ", bins = " << bins << endl;
                continue;
            }
                
            for (int bin = by0; bin == by0 || bin < by1; ++bin) {

                float value = in[bin];

                if (bin == by0 || value > out[y]) {
                    out[y] = value;
                }
            }
        }
    }

    return out;
}