Loading

C# 手动解析灰度PNG图片为Bitmap

问题:

当直接使用文件路径加载8位灰度PNG图片为Bitmap时,Bitmap的格式将会是Format32bppArgb,而不是Format8bppIndexed,这对一些判断会有影响,所以需要手动解析PNG的数据来构造Bitmap

步骤

1. 判断文件格式

若对PNG文件格式不是很了解,阅读本文前可以参考PNG的文件格式 PNG文件格式详解

简而言之,PNG文件头有8个固定字节来标识它,他们是

private static byte[] PNG_IDENTIFIER = { 0x89, 0x50, 0x4E, 0x47, 0x0D, 0x0A, 0x1A, 0x0A };

2. 判断是否为8位灰度图

识别为PNG文件后,需要判断该PNG文件是否为8位的灰度图

在PNG的文件头标识后是PNG文件的第一个数据块IHDR,它的数据域由13个字节组成

域的名称 数据字节数 说明
Width 4 bytes 图像宽度,以像素为单位
Height 4 bytes 图像高度,以像素为单位
Bit depth 1 byte 图像深度:索引彩色图像:1,2,4或8 ;灰度图像:1,2,4,8或16 ;真彩色图像:8或16
ColorType 1 byte 颜色类型:0:灰度图像, 1,2,4,8或16;2:真彩色图像,8或16;3:索引彩色图像,1,2,4或84:带α通道数据的灰度图像,8或16;6:带α通道数据的真彩色图像,8或16
Compression method 1 byte 压缩方法(LZ77派生算法)
Filter method 1 byte 滤波器方法
Interlace method 1 byte 隔行扫描方法:0:非隔行扫描;1: Adam7(由Adam M. Costello开发的7遍隔行扫描方法)

这里我们看颜色深度以及颜色类型就行

 var ihdrData = data[(PNG_IDENTIFIER.Length + 8)..(PNG_IDENTIFIER.Length + 8 + 13)];
 var bitDepth = Convert.ToInt32(ihdrData[8]);
 var colorType = Convert.ToInt32(ihdrData[9]);

这里的data是表示PNG文件的byte数组,+8是因为PNG文件的每个数据块的数据域前都有4个字节的数据域长度和4个字节的数据块类型(名称)


3. 获取全部图像数据块

PNG文件的图像数据由一个或多个图像数据块IDAT构成,并且他们是顺序排列的

这里通过while循环找到所有的IDAT

var compressedSubDats = new List<byte[]>();
var firstDatOffset = FindChunk(data, "IDAT");
var firstDatLength = GetChunkDataLength(data, firstDatOffset);
var firstDat = new byte[firstDatLength];

Array.Copy(data, firstDatOffset + 8, firstDat, 0, firstDatLength);
compressedSubDats.Add(firstDat);

var dataSpan = data.AsSpan().Slice(firstDatOffset + 12 + firstDatLength);
while (Encoding.ASCII.GetString(dataSpan[4..8]) == "IDAT")
{
    var datLength = dataSpan.ReadBinaryInt(0, 4);
    var dat = new byte[datLength];
    dataSpan.Slice(8, datLength).CopyTo(dat);
    compressedSubDats.Add(dat);
    dataSpan = dataSpan.Slice(12 + datLength);
}

var compressedDatLength = compressedSubDats.Sum(a => a.Length);
var compressedDat = new byte[compressedDatLength].AsSpan();
var index = 0;
for (int i = 0; i < compressedSubDats.Count; i++)
{
    var subDat = compressedSubDats[i];
    subDat.CopyTo(compressedDat.Slice(index, subDat.Length));
    index += subDat.Length;
}

4. 解压DAT数据

上一步获得的DAT数据是由Deflate算法压缩后的,我们需要将它解压缩,这里使用.NET自带的DeflateStream进行解压缩

IDAT的数据流以zlib格式存储,结构为

名称 长度
zlib compression method/flags code 1 byte
Additional flags/check bits 1 byte
Compressed data blocks n bytes
Check value 4 bytes

解压缩时去掉前2个字节

var deCompressedDat = MicrosoftDecompress(compressedDat.ToArray()[2..]).AsSpan();
public static byte[] MicrosoftDecompress(byte[] data)
{
    MemoryStream compressed = new MemoryStream(data);
    MemoryStream decompressed = new MemoryStream();
    DeflateStream deflateStream = new DeflateStream(compressed, CompressionMode.Decompress);
    deflateStream.CopyTo(decompressed);
    byte[] result = decompressed.ToArray();
    return result;
}

5. 重建原始数据

PNG的IDAT数据流在压缩前会通过过滤算法将原始数据进行过滤来提高压缩率,这里需要将过滤后的数据进行重建

有关过滤和重建可以参考W3组织的文档

这里定义了一个类来辅助重建

    public class PngFilterByte
    {
        public PngFilterByte(int filterType, int row, int col)
        {
            FilterType = filterType;
            Row = row;
            Column = col;
        }

        public int Row { get; set; }

        public int Column { get; set; }

        public int FilterType { get; set; }

        public PngFilterByte C { get; set; }

        public PngFilterByte B { get; set; }

        public PngFilterByte A { get; set; }

        public int X { get; set; }

        private bool _isTop;

        public bool IsTop
        {
            get => _isTop;
            init
            {
                _isTop = value;
                if (!_isTop) return;
                B = Zero;
            }
        }

        private bool _isLeft;

        public bool IsLeft
        {
            get => _isLeft;
            init
            {
                _isLeft = value;
                if (!_isLeft) return;
                A = Zero;
            }
        }

        public int _filt;

        public int Filt
        {
            get => IsFiltered ? _filt : DoFilter();
            init
            {
                _filt = value;
            }
        }

        public bool IsFiltered { get; set; } = false;

        public int DoFilter()
        {
            _filt = FilterType switch
            {
                0 => X,
                1 => X - A.X,
                2 => X - B.X,
                3 => X - (int)Math.Floor((A.X + B.X) / 2.0M),
                4 => X - Paeth(A.X, B.X, C.X),
                _ => X
            };
            if (_filt > 255) _filt %= 256;
            IsFiltered = true;
            return _filt;
        }

        private int _recon;

        public int Recon
        {
            get => IsReconstructed ? _recon : DoReconstruction();
            init
            {
                _filt = value;
            }
        }

        public bool IsReconstructed { get; set; } = false;

        public int DoReconstruction()
        {
            _recon = FilterType switch
            {
                0 => Filt,
                1 => Filt + A.Recon,
                2 => Filt + B.Recon,
                3 => Filt + (int)Math.Floor((A.Recon + B.Recon) / 2.0M),
                4 => Filt + Paeth(A.Recon, B.Recon, C.Recon),
                _ => Filt
            };
            if (_recon > 255) _recon %= 256;
            X = _recon;
            IsReconstructed = true;
            return _recon;
        }

        private int Paeth(int a, int b, int c)
        {
            var p = a + b - c;
            var pa = Math.Abs(p - a);
            var pb = Math.Abs(p - b);
            var pc = Math.Abs(p - c);
            if (pa <= pb && pa <= pc)
            {
                return a;
            }
            else if (pb <= pc)
            {
                return b;
            }
            else
            {
                return c;
            }
        }

        public static PngFilterByte Zero = new PngFilterByte(0, -1, -1)
        {
            IsFiltered = true,
            IsReconstructed = true,
            X = 0,
            Filt = 0,
            Recon = 0
        };
    }

下面获取重建的数据

首先从IHDR获取宽高

var width = ihdrData.ReadBinaryInt(0, 4);
var height = ihdrData.ReadBinaryInt(4, 4);

按行处理

var filtRowDic = new Dictionary<int, byte[]>();
for (int i = 0; i < height; i++)
{
    var rowData = deCompressedDat.Slice(i * (width + 1), (width + 1));
    filtRowDic.Add(i, rowData.ToArray());
}

var rowColDic = new Dictionary<(int, int), PngFilterByte>();

for (int i = 0; i < height; i++)
{
    var row = filtRowDic[i];
    var filterType = row[0];
    for (int j = 1; j <= width; j++)
    {
        var bt = new PngFilterByte(filterType, i, j - 1)
        {
            Filt = Convert.ToInt32(row[j]),
            IsFiltered = true,
            IsTop = i == 0,
            IsLeft = j == 1
        };
        if (bt.IsTop && bt.IsLeft)
        {
            bt.C=PngFilterByte.Zero;
        }
        if (!bt.IsTop)
        {
            bt.B = rowColDic[(bt.Row - 1, bt.Column)];
        }

        if (!bt.IsLeft)
        {
            bt.A = rowColDic[(bt.Row, bt.Column - 1)];
        }
        rowColDic.Add((bt.Row, bt.Column), bt);
    }
}

var realImageData = new byte[rowColDic.Count];
foreach (var bt in rowColDic.Values)
{
    realImageData[bt.Row * width + bt.Column] = Convert.ToByte(bt.Recon);
}

6. 最后构建灰度Bitmap并赋予数据

using var bitmap = new Bitmap(width, height, PixelFormat.Format8bppIndexed);
ColorPalette cp = bitmap.Palette;
for (int i = 0; i < 256; i++)
{
    cp.Entries[i] = Color.FromArgb(i, i, i);
}
bitmap.Palette = cp;
var bmpData = bitmap.LockBits(new Rectangle(0, 0, width, height), ImageLockMode.ReadWrite, PixelFormat.Format8bppIndexed);
Marshal.Copy(realImageData, 0, bmpData.Scan0, realImageData.Length);
bitmap.UnlockBits(bmpData);

return bitmap;

完整代码

Github Gist


参考:

1. PNG文件格式详解
2. Png的数据解析
3. How to read 8-bit PNG image as 8-bit PNG image only?
4. Portable Network Graphics (PNG) Specification (Second Edition)

posted @ 2023-09-28 14:25  QyQj  阅读(298)  评论(0编辑  收藏  举报