cblkratediststats.java

jpeg2000算法实现

/*
 * CVS identifier:
 *
 * $Id: CBlkRateDistStats.java,v 1.8 2001/02/14 17:34:57 grosbois Exp $
 *
 * Class:                   CBlkRateDistStats
 *
 * Description:             The coded (compressed) code-block with
 *                          rate-distortion statistics.
 *
 *
 *
 * COPYRIGHT:
 * 
 * This software module was originally developed by Rapha雔 Grosbois and
 * Diego Santa Cruz (Swiss Federal Institute of Technology-EPFL); Joel
 * Askel鰂 (Ericsson Radio Systems AB); and Bertrand Berthelot, David
 * Bouchard, F閘ix Henry, Gerard Mozelle and Patrice Onno (Canon Research
 * Centre France S.A) in the course of development of the JPEG2000
 * standard as specified by ISO/IEC 15444 (JPEG 2000 Standard). This
 * software module is an implementation of a part of the JPEG 2000
 * Standard. Swiss Federal Institute of Technology-EPFL, Ericsson Radio
 * Systems AB and Canon Research Centre France S.A (collectively JJ2000
 * Partners) agree not to assert against ISO/IEC and users of the JPEG
 * 2000 Standard (Users) any of their rights under the copyright, not
 * including other intellectual property rights, for this software module
 * with respect to the usage by ISO/IEC and Users of this software module
 * or modifications thereof for use in hardware or software products
 * claiming conformance to the JPEG 2000 Standard. Those intending to use
 * this software module in hardware or software products are advised that
 * their use may infringe existing patents. The original developers of
 * this software module, JJ2000 Partners and ISO/IEC assume no liability
 * for use of this software module or modifications thereof. No license
 * or right to this software module is granted for non JPEG 2000 Standard
 * conforming products. JJ2000 Partners have full right to use this
 * software module for his/her own purpose, assign or donate this
 * software module to any third party and to inhibit third parties from
 * using this software module for non JPEG 2000 Standard conforming
 * products. This copyright notice must be included in all copies or
 * derivative works of this software module.
 * 
 * Copyright (c) 1999/2000 JJ2000 Partners.
 * */
package jj2000.j2k.entropy.encoder;

import jj2000.j2k.entropy.*;
import jj2000.j2k.wavelet.analysis.*;
import java.io.*;

/**
 * This class stores coded (compressed) code-blocks with their associated
 * rate-distortion statistics. This object should always contain all the
 * compressed data of the code-block. It is applicable to the encoder engine
 * only. Some data of the coded-block is stored in the super class, see
 * CodedCBlk.
 *
 * <P>The rate-distortion statistics (i.e. R-D slope) is stored for valid
 * points only. The set of valid points is determined by the entropy coder
 * engine itself. Normally they are selected so as to lye in a convex hull,
 * which can be achived by using the 'selectConvexHull' method of this class,
 * but some other strategies might be employed.
 *
 * <P>The rate (in bytes) for each truncation point (valid or not) is stored
 * in the 'truncRates' array. The rate of a truncation point is the total
 * number of bytes in 'data' (see super class) that have to be decoded to
 * reach the truncation point.
 *
 * <P>The slope (reduction of distortion divided by the increase in rate) at
 * each of the valid truncation points is stored in 'truncSlopes'.
 *
 * <P>The index of each valid truncation point is stored in 'truncIdxs'. The
 * index should be interpreted in the following way: a valid truncation point
 * at position 'n' has the index 'truncIdxs[n]', the rate
 * 'truncRates[truncIdxs[n]]' and the slope 'truncSlopes[n]'. The arrays
 * 'truncIdxs' and 'truncRates' have at least 'nVldTrunc' elements. The
 * 'truncRates' array has at least 'nTotTrunc' elements.
 *
 * <P>In addition the 'isTermPass' array contains a flag for each truncation
 * point (valid and non-valid ones) that tells if the pass is terminated or
 * not. If this variable is null then it means that no pass is terminated,
 * except the last one which always is.
 *
 * <P>The compressed data is stored in the 'data' member variable of the super
 * class.
 *
 * @see CodedCBlk
 * */
public class CBlkRateDistStats extends CodedCBlk { 
    
    /** The subband to which the code-block belongs */
    public SubbandAn sb;

    /** The total number of truncation points */
    public int nTotTrunc;
    
    /** The number of valid truncation points */
    public int nVldTrunc;
    
    /** The rate (in bytes) for each truncation point (valid and non-valid
     * ones) */
    public int truncRates[];
    
    /** The distortion for each truncation point (valid and non-valid ones) */
    public double truncDists[];
    
    /** The negative of the rate-distortion slope for each valid truncation
        point */
    public float truncSlopes[];
    
    /** The indices of the valid truncation points, in increasing
     * order. */
    public int truncIdxs[];
    
    /** Array of flags indicating terminated passes (valid or non-valid
     * truncation points). */
    public boolean isTermPass[];

    /** The number of ROI coefficients in the code-block */
    public int nROIcoeff = 0;

    /** Number of ROI coding passes */
    public int nROIcp = 0;

    /**
     * Creates a new CBlkRateDistStats object without allocating any space for
     * 'truncRates', 'truncSlopes', 'truncDists' and 'truncIdxs' or 'data'.
     * */
    public CBlkRateDistStats() {
    }

    /**
     * Creates a new CBlkRateDistStats object and initializes the valid
     * truncation points, their rates and their slopes, from the 'rates' and
     * 'dist' arrays. The 'rates', 'dist' and 'termp' arrays must contain the
     * rate (in bytes), the reduction in distortion (from nothing coded) and
     * the flag indicating if termination is used, respectively, for each
     * truncation point.
     *
     * <P>The valid truncation points are selected by taking them as lying on
     * a convex hull. This is done by calling the method selectConvexHull().
     *
     * <P> Note that the arrays 'rates' and 'termp' are copied, not
     * referenced, so they can be modified after a call to this constructor.
     *
     * @param m The horizontal index of the code-block, within the subband.
     *
     * @param n The vertical index of the code-block, within the subband.
     *
     * @param skipMSBP The number of skipped most significant bit-planes for
     * this code-block.
     *
     * @param data The compressed data. This array is referenced by this
     * object so it should not be modified after.
     *
     * @param rates The rates (in bytes) for each truncation point in the
     * compressed data. This array is modified by the method but no reference
     * is kept to it.
     *
     * @param dists The reduction in distortion (with respect to no information 
     * coded) for each truncation point. This array is modified by the method
     * but no reference is kept to it.
     *
     * @param termp An array of boolean flags indicating, for each pass, if a
     * pass is terminated or not (true if terminated). If null then it is
     * assumed that no pass is terminated except the last one which always is.
     * 
     * @param np The number of truncation points contained in 'rates', 'dist'
     * and 'termp'.
     *
     * @param inclast If false the convex hull is constructed as for lossy
     * coding. If true it is constructed as for lossless coding, in which case
     * it is ensured that all bit-planes are sent (i.e. the last truncation
     * point is always included).
     * */
    public CBlkRateDistStats(int m, int n, int skipMSBP, byte data[],
                          int rates[], double dists[], boolean termp[],
                          int np, boolean inclast) {
        super(m,n,skipMSBP,data);
        selectConvexHull(rates,dists,termp,np,inclast);
    }
    
    /**
     * Compute the rate-distorsion slopes and selects those that lie in a
     * convex hull. It will compute the slopes, select the ones that form the
     * convex hull and initialize the 'truncIdxs' and 'truncSlopes' arrays, as
     * well as 'nVldTrunc', with the selected truncation points. It will also
     * initialize 'truncRates' and 'isTermPass' arrays, as well as
     * 'nTotTrunc', with all the truncation points (selected or not).
     *
     * <P> Note that the arrays 'rates' and 'termp' are copied, not
     * referenced, so they can be modified after a call to this method.
     *
     * @param rates The rates (in bytes) for each truncation point in the
     * compressed data. This array is modified by the method.
     *
     * @param dists The reduction in distortion (with respect to no
     * information coded) for each truncation point. This array is modified by
     * the method.
     *
     * @param termp An array of boolean flags indicating, for each pass, if a
     * pass is terminated or not (true if terminated). If null then it is
     * assumed that no pass is terminated except the last one which always is.
     * 
     * @param n The number of truncation points contained in 'rates', 'dist'
     * and 'termp'.
     *
     * @param inclast If false the convex hull is constructed as for lossy
     * coding. If true it is constructed as for lossless coding, in which case
     * it is ensured that all bit-planes are sent (i.e. the last truncation
     * point is always included).
     * */
    public void selectConvexHull(int rates[], double dists[], boolean termp[],
                                 int n, boolean inclast) {
        int first_pnt;    // The first point containing some coded data
        int p;            // last selected point
        int k;            // current point
        int i;            // current valid point
        int npnt;         // number of selected (i.e. valid) points
        int delta_rate;   // Rate difference
        double delta_dist; // Distortion difference
        float k_slope;    // R-D slope for the current point
        float p_slope;    // R-D slope for the last selected point
        int ll_rate;      // Rate for "lossless" coding (i.e. all coded info)

        // Convention: when a negative value is stored in 'rates' it meas an
        // invalid point. The absolute value is always the rate for that point.

        // Look for first point with some coded info (rate not 0)
        first_pnt = 0;
        while (first_pnt < n && rates[first_pnt] <= 0) {
            first_pnt++;
        }

        // Select the valid points
        npnt = n-first_pnt;
        p_slope = 0f; // To keep compiler happy
ploop:
        do {
            p = -1;
            for (k=first_pnt; k<n; k++) {
                if (rates[k] < 0) { // Already invalidated point
                    continue;
                }
                // Calculate decrease in distortion and rate
                if (p >= 0) {
                    delta_rate = rates[k]-rates[p];
                    delta_dist = dists[k]-dists[p];
                }
                else { // This is with respect to no info coded
                    delta_rate = rates[k];
                    delta_dist = dists[k];
                }
                // If exactly same distortion don't eliminate if the rates are 
                // equal, otherwise it can lead to infinite slope in lossless
                // coding.
                if (delta_dist < 0f || (delta_dist == 0f && delta_rate > 0)) {
                    // This point increases distortion => invalidate
                    rates[k] = -rates[k];
                    npnt--;
                    continue; // Goto next point
                }
                k_slope = (float)(delta_dist/delta_rate);
                // Check that there is a decrease in distortion, slope is not
                // infinite (i.e. delta_dist is not 0) and slope is
                // decreasing.
                if (p>=0 &&
                    (delta_rate <= 0 || k_slope >= p_slope )) {
                    // Last point was not good
                    rates[p] = -rates[p]; // Remove p from valid points
                    npnt--;
                    continue ploop; // Restart from the first one
                }
                else {
                    p_slope = k_slope;
                    p = k;
                }
            }
            // If we get to last point we are done
            break;
        } while (true); // We end the loop with the break statement

        // If in lossless mode make sure we don't eliminate any last bit-planes 
        // from being sent.
        if (inclast && n > 0 && rates[n-1] < 0) {
            rates[n-1] = -rates[n-1];
            // This rate can never be equal to any previous selected rate,
            // given the selection algorithm above, so no problem arises of
            // infinite slopes.
            npnt++;
        }

        // Initialize the arrays of this object
        nTotTrunc = n;
        nVldTrunc = npnt;
        truncRates = new int[n];
        truncDists = new double[n];
        truncSlopes = new float[npnt];
        truncIdxs = new int[npnt];
        if (termp != null) {
            isTermPass = new boolean[n];
            System.arraycopy(termp,0,isTermPass,0,n);
        }
        else {
            isTermPass = null;
        }
        System.arraycopy(rates,0,truncRates,0,n);
        for (k=first_pnt, p=-1, i=0; k<n; k++) {
            if (rates[k]>0) { // A valid point
                truncDists[k] = dists[k];
                if (p<0) { // Only arrives at first valid point
                    truncSlopes[i] = (float)(dists[k]/rates[k]);
                }
                else {
                    truncSlopes[i] = (float)((dists[k]-dists[p])/
                                             (rates[k]-rates[p]));
                }
                truncIdxs[i] = k;
                i++;
                p = k;
            }
            else {
                truncDists[k] = -1;
                truncRates[k] = -truncRates[k];
            }
        }
    }
       
    /**
     * Returns the contents of the object in a string. This is used for
     * debugging.
     *
     * @return A string with the contents of the object
     * */
    public String toString() {
        return super.toString() +
            "\n nVldTrunc = "+nVldTrunc+", nTotTrunc="+nTotTrunc+", num. ROI"+
            " coeff="+nROIcoeff+", num. ROI coding passes="+nROIcp;
    }
}