/*
 *  This file is part of Aptdec.
 *  Copyright (c) 2004-2009 Thierry Leconte (F4DWV), Xerbo (xerbo@protonmail.com) 2019-2022
 *
 *  Aptdec 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.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program.  If not, see <https://www.gnu.org/licenses/>.
 *
 */

#include "image.h"

#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#include "algebra.h"
#include "apt.h"
#include "util.h"

static linear_t compute_regression(float *wedges) {
    //				    { 0.106, 0.215, 0.324, 0.433, 0.542,  0.652, 0.78,   0.87,  0.0 }
    const float teleramp[9] = {31.07, 63.02, 94.96, 126.9, 158.86, 191.1, 228.62, 255.0, 0.0};

    return linear_regression(wedges, teleramp, 9);
}

static float tele[16];
static float Cs;

void apt_histogramEqualise(float **prow, int nrow, int offset, int width) {
    // Plot histogram
    int histogram[256] = {0};
    for (int y = 0; y < nrow; y++)
        for (int x = 0; x < width; x++) histogram[(int)CLIP(prow[y][x + offset], 0, 255)]++;

    // Calculate cumulative frequency
    long sum = 0, cf[256] = {0};
    for (int i = 0; i < 255; i++) {
        sum += histogram[i];
        cf[i] = sum;
    }

    // Apply histogram
    int area = nrow * width;
    for (int y = 0; y < nrow; y++) {
        for (int x = 0; x < width; x++) {
            int k = (int)prow[y][x + offset];
            prow[y][x + offset] = (256.0f / area) * cf[k];
        }
    }
}

void apt_linearEnhance(float **prow, int nrow, int offset, int width) {
    // Plot histogram
    int histogram[256] = {0};
    for (int y = 0; y < nrow; y++)
        for (int x = 0; x < width; x++) histogram[(int)CLIP(prow[y][x + offset], 0, 255)]++;

    // Find min/max points
    int min = -1, max = -1;
    for (int i = 5; i < 250; i++) {
        if (histogram[i] / width / (nrow / 255.0) > 0.1) {
            if (min == -1) min = i;
            max = i;
        }
    }

    // Stretch the brightness into the new range
    for (int y = 0; y < nrow; y++) {
        for (int x = 0; x < width; x++) {
            prow[y][x + offset] = (prow[y][x + offset] - min) / (max - min) * 255.0f;
            prow[y][x + offset] = CLIP(prow[y][x + offset], 0.0f, 255.0f);
        }
    }
}

// Brightness calibrate, including telemetry
void calibrateImage(float **prow, int nrow, int offset, int width, linear_t regr) {
    offset -= APT_SYNC_WIDTH + APT_SPC_WIDTH;

    for (int y = 0; y < nrow; y++) {
        for (int x = 0; x < width + APT_SYNC_WIDTH + APT_SPC_WIDTH + APT_TELE_WIDTH; x++) {
            float pv = linear_calc(prow[y][x + offset], regr);
            prow[y][x + offset] = CLIP(pv, 0, 255);
        }
    }
}

float teleNoise(float *wedges) {
    float pattern[9] = {31.07, 63.02, 94.96, 126.9, 158.86, 191.1, 228.62, 255.0, 0.0};
    float noise = 0;
    for (int i = 0; i < 9; i++) noise += fabs(wedges[i] - pattern[i]);

    return noise;
}

// Get telemetry data for thermal calibration
apt_channel_t apt_calibrate(float **prow, int nrow, int offset, int width) {
    float teleline[APT_MAX_HEIGHT] = {0.0};
    float wedge[16];
    linear_t regr[APT_MAX_HEIGHT / APT_FRAME_LEN + 1];
    int telestart, mtelestart = 0;
    int channel = -1;

    // The minimum rows required to decode a full frame
    if (nrow < APT_CALIBRATION_ROWS) {
        error_noexit("Telemetry decoding error, not enough rows");
        return APT_CHANNEL_UNKNOWN;
    }

    // Calculate average of a row of telemetry
    for (int y = 0; y < nrow; y++) {
        for (int x = 3; x < 43; x++) teleline[y] += prow[y][x + offset + width];

        teleline[y] /= 40.0;
    }

    /* Wedge 7 is white and 8 is black, this will have the largest
     * difference in brightness, this can be used to find the current
     * position within the telemetry.
     */
    float max = 0.0f;
    for (int n = nrow / 3 - 64; n < 2 * nrow / 3 - 64; n++) {
        float df;

        // (sum 4px below) - (sum 4px above)
        df = (float)((teleline[n - 4] + teleline[n - 3] + teleline[n - 2] + teleline[n - 1]) -
                     (teleline[n + 0] + teleline[n + 1] + teleline[n + 2] + teleline[n + 3]));

        // Find the maximum difference
        if (df > max) {
            mtelestart = n;
            max = df;
        }
    }

    telestart = (mtelestart + 64) % APT_FRAME_LEN;

    // Make sure that theres at least one full frame in the image
    if (nrow < telestart + APT_FRAME_LEN) {
        error_noexit("Telemetry decoding error, not enough rows");
        return APT_CHANNEL_UNKNOWN;
    }

    // Find the least noisy frame
    float minNoise = -1;
    int bestFrame = -1;
    for (int n = telestart, k = 0; n < nrow - APT_FRAME_LEN; n += APT_FRAME_LEN, k++) {
        int j;

        for (j = 0; j < 16; j++) {
            int i;

            wedge[j] = 0.0;
            for (i = 1; i < 7; i++) wedge[j] += teleline[n + j * 8 + i];
            wedge[j] /= 6;
        }

        float noise = teleNoise(wedge);
        if (noise < minNoise || minNoise == -1) {
            minNoise = noise;
            bestFrame = k;

            // Compute & apply regression on the wedges
            regr[k] = compute_regression(wedge);
            for (int j = 0; j < 16; j++) tele[j] = linear_calc(wedge[j], regr[k]);

            /* Compare the channel ID wedge to the reference
             * wedges, the wedge with the closest match will
             * be the channel ID
             */
            float min = -1;
            for (int j = 0; j < 6; j++) {
                float df = (float)(tele[15] - tele[j]);
                df *= df;

                if (df < min || min == -1) {
                    channel = j;
                    min = df;
                }
            }

            // Find the brightness of the minute marker, I don't really know what for
            Cs = 0.0;
            int i, j = n;
            for (i = 0, j = n; j < n + APT_FRAME_LEN; j++) {
                float csline = 0.0;
                for (int l = 3; l < 43; l++) csline += prow[n][l + offset - APT_SPC_WIDTH];
                csline /= 40.0;

                if (csline > 50.0) {
                    Cs += csline;
                    i++;
                }
            }
            Cs = linear_calc((Cs / i), regr[k]);
        }
    }

    if (bestFrame == -1) {
        error_noexit("Something has gone very wrong, please file a bug report");
        return APT_CHANNEL_UNKNOWN;
    }

    calibrateImage(prow, nrow, offset, width, regr[bestFrame]);

    return (apt_channel_t)(channel + 1);
}

extern float quick_select(float arr[], int n);

// Biased median denoise, pretyt ugly
#define TRIG_LEVEL 40
void apt_denoise(float **prow, int nrow, int offset, int width) {
    for (int y = 2; y < nrow - 2; y++) {
        for (int x = offset + 1; x < offset + width - 1; x++) {
            if (prow[y][x + 1] - prow[y][x] > TRIG_LEVEL || prow[y][x - 1] - prow[y][x] > TRIG_LEVEL ||
                prow[y + 1][x] - prow[y][x] > TRIG_LEVEL || prow[y - 1][x] - prow[y][x] > TRIG_LEVEL) {
                prow[y][x] = quick_select((float[]){prow[y + 2][x - 1], prow[y + 2][x], prow[y + 2][x + 1], prow[y + 1][x - 1],
                                                    prow[y + 1][x], prow[y + 1][x + 1], prow[y - 1][x - 1], prow[y - 1][x],
                                                    prow[y - 1][x + 1], prow[y - 2][x - 1], prow[y - 2][x], prow[y - 2][x + 1]},
                                          12);
            }
        }
    }
}
#undef TRIG_LEVEL

// Flips a channel, for northbound passes
void apt_flipImage(apt_image_t *img, int width, int offset) {
    for (int y = 1; y < img->nrow; y++) {
        for (int x = 1; x < ceil(width / 2.0); x++) {
            // Flip top-left & bottom-right
            float buffer = img->prow[img->nrow - y][offset + x];
            img->prow[img->nrow - y][offset + x] = img->prow[y][offset + (width - x)];
            img->prow[y][offset + (width - x)] = buffer;
        }
    }
}

// Calculate crop to reomve noise from the start and end of an image
int apt_cropNoise(apt_image_t *img) {
#define NOISE_THRESH 180.0

    // Average value of minute marker
    float spc_rows[APT_MAX_HEIGHT] = {0.0};
    int startCrop = 0;
    int endCrop = img->nrow;
    for (int y = 0; y < img->nrow; y++) {
        for (int x = 0; x < APT_SPC_WIDTH; x++) {
            spc_rows[y] += img->prow[y][x + (APT_CHB_OFFSET - APT_SPC_WIDTH)];
        }
        spc_rows[y] /= APT_SPC_WIDTH;

        // Skip minute markings
        if (spc_rows[y] < 10) {
            spc_rows[y] = spc_rows[y - 1];
        }
    }

    // 3 row average
    for (int y = 0; y < img->nrow; y++) {
        spc_rows[y] = (spc_rows[y + 1] + spc_rows[y + 2] + spc_rows[y + 3]) / 3;
        // img.prow[y][0] = spc_rows[y];
    }

    // Find ends
    for (int y = 0; y < img->nrow - 1; y++) {
        if (spc_rows[y] > NOISE_THRESH) {
            endCrop = y;
        }
    }
    for (int y = img->nrow; y > 0; y--) {
        if (spc_rows[y] > NOISE_THRESH) {
            startCrop = y;
        }
    }

    // printf("Crop rows: %i -> %i\n", startCrop, endCrop);

    // Remove the noisy rows at start
    for (int y = 0; y < img->nrow - startCrop; y++) {
        memmove(img->prow[y], img->prow[y + startCrop], sizeof(float) * APT_PROW_WIDTH);
    }

    // Ignore the noisy rows at the end
    img->nrow = (endCrop - startCrop);

    return startCrop;
}

// --- Visible and Temperature Calibration --- //
#include "calibration.h"

typedef struct {
    float Nbb;
    float Cs;
    float Cb;
    int ch;
} tempparam_t;

// IR channel temperature compensation
tempparam_t tempcomp(float t[16], int ch, int satnum) {
    tempparam_t tpr;
    tpr.ch = ch - 4;

    const calibration_t calibration = get_calibration(satnum);
    const float Vc = calibration.rad[tpr.ch].vc;
    const float A = calibration.rad[tpr.ch].A;
    const float B = calibration.rad[tpr.ch].B;

    // Compute PRT temperature
    float T[4];
    for (size_t n = 0; n < 4; n++) {
        T[n] = quadratic_calc(t[n + 9] * 4.0, calibration.prt[n]);
    }

    float Tbb = (T[0] + T[1] + T[2] + T[3]) / 4.0;  // Blackbody temperature
    float Tbbstar = A + Tbb * B;                    // Effective blackbody temperature

    tpr.Nbb = C1 * pow(Vc, 3) / (expf(C2 * Vc / Tbbstar) - 1.0f);  // Blackbody radiance

    tpr.Cs = 246.4 * 4.0;  // FIXME
    tpr.Cb = t[14] * 4.0;
    return tpr;
}

// IR channel temperature calibration
static float tempcal(float Ce, int satnum, tempparam_t tpr) {
    const calibration_t calibration = get_calibration(satnum);
    const float Ns = calibration.cor[tpr.ch].Ns;
    const float Vc = calibration.rad[tpr.ch].vc;
    const float A = calibration.rad[tpr.ch].A;
    const float B = calibration.rad[tpr.ch].B;

    float Nl = Ns + (tpr.Nbb - Ns) * (tpr.Cs - Ce * 4.0) / (tpr.Cs - tpr.Cb);  // Linear radiance estimate
    float Nc = quadratic_calc(Nl, calibration.cor[tpr.ch].quadratic);          // Non-linear correction
    float Ne = Nl + Nc;                                                        // Corrected radiance

    float Testar = C2 * Vc / logf(C1 * powf(Vc, 3) / Ne + 1.0);  // Equivlent black body temperature
    float Te = (Testar - A) / B;                                 // Temperature (kelvin)

    // Convert to celsius
    Te -= 273.15;
    // Rescale to 0-255 for -100°C to +60°C
    return (Te + 100.0) / 160.0 * 255.0;
}

// Temperature calibration wrapper
void apt_calibrate_thermal(int satnum, apt_image_t *img, int offset, int width) {
    tempparam_t temp = tempcomp(tele, img->chB, satnum);

    for (int y = 0; y < img->nrow; y++) {
        for (int x = 0; x < width; x++) {
            img->prow[y][x + offset] = (float)tempcal(img->prow[y][x + offset], satnum, temp);
        }
    }
}

float calibrate_pixel(float value, int channel, calibration_t cal) {
    if (value > cal.visible[channel].cutoff) {
        return linear_calc(value * 4.0f, cal.visible[channel].high) * 255.0f / 100.0f;
    } else {
        return linear_calc(value * 4.0f, cal.visible[channel].low) * 255.0f / 100.0f;
    }
}

void apt_calibrate_visible(int satnum, apt_image_t *img, int offset, int width) {
    const calibration_t calibration = get_calibration(satnum);
    int channel = img->chA - 1;

    for (int y = 0; y < img->nrow; y++) {
        for (int x = 0; x < width; x++) {
            img->prow[y][x + offset] = clamp(calibrate_pixel(img->prow[y][x + offset], channel, calibration), 255.0f, 0.0f);
        }
    }
}