path: root/src/plugins/graphicssystems/meego/dithering.cpp

/****************************************************************************
**
** Copyright (C) 2010 Nokia Corporation and/or its subsidiary(-ies).
** All rights reserved.
** Contact: Nokia Corporation (qt-info@nokia.com)
**
** This file is part of the plugins of the Qt Toolkit.
**
** $QT_BEGIN_LICENSE:LGPL$
** No Commercial Usage
** This file contains pre-release code and may not be distributed.
** You may use this file in accordance with the terms and conditions
** contained in the Technology Preview License Agreement accompanying
** this package.
**
** GNU Lesser General Public License Usage
** Alternatively, this file may be used under the terms of the GNU Lesser
** General Public License version 2.1 as published by the Free Software
** Foundation and appearing in the file LICENSE.LGPL included in the
** packaging of this file.  Please review the following information to
** ensure the GNU Lesser General Public License version 2.1 requirements
** will be met: http://www.gnu.org/licenses/old-licenses/lgpl-2.1.html.
**
** In addition, as a special exception, Nokia gives you certain additional
** rights.  These rights are described in the Nokia Qt LGPL Exception
** version 1.1, included in the file LGPL_EXCEPTION.txt in this package.
**
** If you have questions regarding the use of this file, please contact
** Nokia at qt-info@nokia.com.
**
**
**
**
**
**
**
**
** $QT_END_LICENSE$
**
****************************************************************************/

// This is an implementation of the 32bit => 16bit Floyd-Steinberg dithering.
// The alghorithm used here is not the fastest possible but it's prolly fast enough:
// uses look-up tables, integer-only arthmetics and works in one pass on two lines
// at a time. It's a high-quality dithering using 1/8 diffusion precission.
// Two functions here to look at:
//
//   * convertRGBA32_to_RGB565
//   * convertRGBA32_to_RGBA4444
//
// Each channel (RGBA) is diffused independently and alpha is dithered too.

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

// Gets a component (red = 1, green = 2...) from a RGBA data structure.
// data is unsigned char. stride is the number of bytes per line.
#define GET_RGBA_COMPONENT(data, x, y, stride, c) (data[(y * stride) + (x << 2) + c])

// Writes a new pixel with r, g, b to data in 565 16bit format. Data is a short.
#define PUT_565(data, x, y, width, r, g, b) (data[(y * width) + x] = (r << 11) | (g << 5) | b)

// Writes a new pixel with r, g, b, a to data in 4444 RGBA 16bit format. Data is a short.
#define PUT_4444(data, x, y, width, r, g, b, a) (data[(y * width) + x] = (r << 12) | (g << 8) | (b << 4) | a)

// Writes(ads) a new value to the diffusion accumulator. accumulator is a short.
// x, y is a position in the accumulation buffer. y can be 0 or 1 -- we operate on two lines at time.
#define ACCUMULATE(accumulator, x, y, width, v) if (x < width && x > 0) accumulator[(y * width) + x] += v

// Clamps a value to be in 0..255 range.
#define CLAMP_256(v) if (v > 255) v = 255; if (v < 0) v = 0;

// Converts incoming RGB32 (QImage::Format_RGB32) to RGB565. Returns the newly allocated data.
unsigned short* convertRGB32_to_RGB565(const unsigned char *in, int width, int height, int stride)
{
    // Output line stride. Alligned to 4 bytes.
    int alignedWidth = width;
    if (alignedWidth % 2 > 0)
        alignedWidth++;

    // Will store output
    unsigned short *out = (unsigned short *) malloc(alignedWidth * height * 2);

    // Lookup tables for the 8bit => 6bit and 8bit => 5bit conversion
    unsigned char lookup_8bit_to_5bit[256];
    short lookup_8bit_to_5bit_diff[256];
    unsigned char lookup_8bit_to_6bit[256];
    short lookup_8bit_to_6bit_diff[256];

    // Macros for the conversion using the lookup table.
    #define CONVERT_8BIT_TO_5BIT(v) (lookup_8bit_to_5bit[v])
    #define DIFF_8BIT_TO_5BIT(v) (lookup_8bit_to_5bit_diff[v])

    #define CONVERT_8BIT_TO_6BIT(v) (lookup_8bit_to_6bit[v])
    #define DIFF_8BIT_TO_6BIT(v) (lookup_8bit_to_6bit_diff[v])

    int i;
    int x, y, c; // Pixel we're processing. c is component number (0, 1, 2 for r, b, b)
    short component[3]; // Stores the new components (r, g, b) for pixel produced during conversion
    short diff; // The difference between the converted value and the original one. To be accumulated.
    QVarLengthArray <short> accumulatorData(3 * width * 2); // Data for three acumulators for r, g, b. Each accumulator is two lines.
    short *accumulator[3]; // Helper for accessing the accumulator on a per-channel basis more easily.
    accumulator[0] = accumulatorData.data();
    accumulator[1] = accumulatorData.data() + width;
    accumulator[2] = accumulatorData.data() + (width * 2);

    // Produce the conversion lookup tables.
    for (i = 0; i < 256; i++) {
        lookup_8bit_to_5bit[i] = round(i / 8.0);
        if (lookup_8bit_to_5bit[i] > 31)
            lookup_8bit_to_5bit[i] -= 1;

        // Before bitshifts: (i * 8) - (... * 8 * 8)
        lookup_8bit_to_5bit_diff[i] = (i << 3) - (lookup_8bit_to_5bit[i] << 6);

        lookup_8bit_to_6bit[i] = round(i / 4.0);
        if (lookup_8bit_to_6bit[i] > 63)
            lookup_8bit_to_6bit[i] -= 1;

        // Before bitshifts: (i * 8) - (... * 4 * 8)
        lookup_8bit_to_6bit_diff[i] = (i << 3) - (lookup_8bit_to_6bit[i] << 5);
    }

    // Clear the accumulators
    memset(accumulator[0], 0, width * 4);
    memset(accumulator[1], 0, width * 4);
    memset(accumulator[2], 0, width * 4);

    // For each line...
    for (y = 0; y < height; y++) {

        // For each accumulator, move the second line (index 1) to replace the first line (index 0).
        // Clear the second line (index 1)
        memcpy(accumulator[0], accumulator[0] + width, width * 2);
        memset(accumulator[0] + width, 0, width * 2);

        memcpy(accumulator[1], accumulator[1] + width, width * 2);
        memset(accumulator[1] + width, 0, width * 2);

        memcpy(accumulator[2], accumulator[2] + width, width * 2);
        memset(accumulator[2] + width, 0, width * 2);

        // For each column....
        for (x = 0; x < width; x++) {

            // For each component (r, g, b)...
            for (c = 0; c < 3; c++) {

                // Get the 8bit value from the original image
                component[c] = GET_RGBA_COMPONENT(in, x, y, stride, c);

                // Add the diffusion for this pixel we stored in the accumulator.
                // >> 7 because the values in accumulator are stored * 128
                component[c] += accumulator[c][x] >> 7;

                // Make sure we're not over the boundaries.
                CLAMP_256(component[c]);

                // For green component we use 6 bits. Otherwise 5 bits.
                // Store the difference from converting 8bit => 6 bit and the orig pixel.
                // Convert 8bit => 6(5) bit.
                if (c == 1) {
                    diff = DIFF_8BIT_TO_6BIT(component[c]);
                    component[c] = CONVERT_8BIT_TO_6BIT(component[c]);
                } else {
                    diff = DIFF_8BIT_TO_5BIT(component[c]);
                    component[c] = CONVERT_8BIT_TO_5BIT(component[c]);
                }

                // Distribute the difference according to the matrix in the
                // accumulation bufffer.
                ACCUMULATE(accumulator[c], x + 1, 0, width, diff * 7);
                ACCUMULATE(accumulator[c], x - 1, 1, width, diff * 3);
                ACCUMULATE(accumulator[c], x, 1, width, diff * 5);
                ACCUMULATE(accumulator[c], x + 1, 1, width, diff * 1);
            }

            // Write the newly produced pixel
            PUT_565(out, x, y, alignedWidth, component[2], component[1], component[0]);
        }
    }

    return out;
}

// Converts incoming RGBA32 (QImage::Format_ARGB32_Premultiplied) to RGB565. Returns the newly allocated data.
// This function is similar (yet different) to the _565 variant but it makes sense to duplicate it here for simplicity.
// The output has each scan line aligned to 4 bytes (as expected by GL by default).
unsigned short* convertARGB32_to_RGBA4444(const unsigned char *in, int width, int height, int stride)
{
    // Output line stride. Alligned to 4 bytes.
    int alignedWidth = width;
    if (alignedWidth % 2 > 0)
        alignedWidth++;

    // Will store output
    unsigned short *out = (unsigned short *) malloc(alignedWidth * 2 * height);

    // Lookup tables for the 8bit => 4bit conversion
    unsigned char lookup_8bit_to_4bit[256];
    short lookup_8bit_to_4bit_diff[256];

    // Macros for the conversion using the lookup table.
    #define CONVERT_8BIT_TO_4BIT(v) (lookup_8bit_to_4bit[v])
    #define DIFF_8BIT_TO_4BIT(v) (lookup_8bit_to_4bit_diff[v])

    int i;
    int x, y, c; // Pixel we're processing. c is component number (0, 1, 2, 3 for r, b, b, a)
    short component[4]; // Stores the new components (r, g, b, a) for pixel produced during conversion
    short diff; // The difference between the converted value and the original one. To be accumulated.
    QVarLengthArray <short> accumulatorData(4 * width * 2); // Data for three acumulators for r, g, b. Each accumulator is two lines.
    short *accumulator[4]; // Helper for accessing the accumulator on a per-channel basis more easily.
    accumulator[0] = accumulatorData.data();
    accumulator[1] = accumulatorData.data() + width;
    accumulator[2] = accumulatorData.data() + (width * 2);
    accumulator[3] = accumulatorData.data() + (width * 3);

    // Produce the conversion lookup tables.
    for (i = 0; i < 256; i++) {
        lookup_8bit_to_4bit[i] = round(i / 16.0);
        if (lookup_8bit_to_4bit[i] > 15)
            lookup_8bit_to_4bit[i] -= 1;

        // Before bitshifts: (i * 8) - (... * 16 * 8)
        lookup_8bit_to_4bit_diff[i] = (i << 3) - (lookup_8bit_to_4bit[i] << 7);
    }

    // Clear the accumulators
    memset(accumulator[0], 0, width * 4);
    memset(accumulator[1], 0, width * 4);
    memset(accumulator[2], 0, width * 4);
    memset(accumulator[3], 0, width * 4);

    // For each line...
    for (y = 0; y < height; y++) {

        // For each component (r, g, b, a)...
        memcpy(accumulator[0], accumulator[0] + width, width * 2);
        memset(accumulator[0] + width, 0, width * 2);

        memcpy(accumulator[1], accumulator[1] + width, width * 2);
        memset(accumulator[1] + width, 0, width * 2);

        memcpy(accumulator[2], accumulator[2] + width, width * 2);
        memset(accumulator[2] + width, 0, width * 2);

        memcpy(accumulator[3], accumulator[3] + width, width * 2);
        memset(accumulator[3] + width, 0, width * 2);

        // For each column....
        for (x = 0; x < width; x++) {

            // For each component (r, g, b, a)...
            for (c = 0; c < 4; c++) {

                // Get the 8bit value from the original image
                component[c] = GET_RGBA_COMPONENT(in, x, y, stride, c);

                // Add the diffusion for this pixel we stored in the accumulator.
                // >> 7 because the values in accumulator are stored * 128
                component[c] += accumulator[c][x] >> 7;

                // Make sure we're not over the boundaries.
                CLAMP_256(component[c]);

                // Store the difference from converting 8bit => 4bit and the orig pixel.
                // Convert 8bit => 4bit.
                diff = DIFF_8BIT_TO_4BIT(component[c]);
                component[c] = CONVERT_8BIT_TO_4BIT(component[c]);

                // Distribute the difference according to the matrix in the
                // accumulation bufffer.
                ACCUMULATE(accumulator[c], x + 1, 0, width, diff * 7);
                ACCUMULATE(accumulator[c], x - 1, 1, width, diff * 3);
                ACCUMULATE(accumulator[c], x, 1, width, diff * 5);
                ACCUMULATE(accumulator[c], x + 1, 1, width, diff * 1);
            }

            // Write the newly produced pixel
            PUT_4444(out, x, y, alignedWidth, component[0], component[1], component[2], component[3]);
        }
    }

    return out;
}