✨ feat(AudioImage/ImageAudio): 将音频与图像互转改为 RGBA 字节直接映射，实现可逆转换，并检查输出文件是否已存在 (51532752) · Commits · shRabbit / TestV · GitLab

TestV/Playgrounds/AudioImage.cs

+62 −53

Original line number	Diff line number	Diff line
		@@ -14,10 +14,9 @@ public class AudioImage : IPlayground
		{
		public string LaunchCmd { get; } = "ai";
		public string? Name { get; } = "Audio Image";
		public string? Description { get; } = "将音频采样反向变换为图片（默认PNG，可选JPG/BMP，可设置宽度或高度）";
		public string? Description { get; } = "将音频采样反向变换为图片（可逆转换，默认PNG，可选JPG/BMP，可设置宽度或高度）";

		private const int BitsPerPixel = 32; // 与 ImageAudio 对应：R/G/B/A 各 8 bit
		private const int FramesPerPixel = BitsPerPixel;
		private const int Channels = 2; // 立体声（与 ImageAudio 对应）

		public async Task Run(string[] args)
		{
		@@ -31,6 +30,12 @@ public class AudioImage : IPlayground
		return;
		}

		if (File.Exists(opts.OutputPath))
		{
		AnsiConsole.MarkupLine($"[red]输出文件已存在: {opts.OutputPath}[/]");
		return;
		}

		await Task.Run(() => Decode(opts));
		}

		@@ -46,7 +51,6 @@ public class AudioImage : IPlayground
		private static Options? ParseOptions(string[] args)
		{
		var audioPath = args[0];
		// 默认输出路径：与音频同目录，扩展名 png
		string outPath = Path.ChangeExtension(audioPath, ".png");
		var opts = new Options { AudioPath = audioPath, OutputPath = outPath };

		@@ -78,7 +82,6 @@ public class AudioImage : IPlayground
		}
		}

		// 若未显式指定 --format，按输出扩展名推断
		if (!IsFormatArg(args))
		{
		var ext = Path.GetExtension(opts.OutputPath).TrimStart('.').ToLowerInvariant();
		@@ -99,15 +102,17 @@ public class AudioImage : IPlayground
		{
		using var reader = new AudioFileReader(opts.AudioPath);
		int audioChannels = reader.WaveFormat.Channels;
		if (audioChannels < 1)

		if (audioChannels != 2)
		{
		AnsiConsole.MarkupLine("[red]无效的音频声道数。[/]");
		AnsiConsole.MarkupLine($"[red]仅支持立体声音频（当前: {audioChannels} 声道）。ImageAudio 生成的音频为立体声。[/]");
		return;
		}

		long totalSamples = reader.Length / sizeof(float); // float 采样数
		long totalFrames = totalSamples / audioChannels;
		long totalPixels = totalFrames / FramesPerPixel;
		// 每2个float采样=1个立体声帧=2个像素
		long totalSamples = reader.Length / sizeof(float);
		long totalFrames = totalSamples / Channels;
		long totalPixels = totalFrames * 2;

		if (totalPixels <= 0)
		{
		@@ -124,28 +129,51 @@ public class AudioImage : IPlayground
		(usedPixels < totalPixels ? $" [yellow](丢弃 {totalPixels - usedPixels:N0} 像素采样)[/]" : "") + "[/]");

		using var image = new Image<Rgba32>(width, height);
		var rowBuf = new float[width * FramesPerPixel * audioChannels];
		long decodedPixels = 0;
		var sampleBuf = new float[8192];
		int bufPos = 0;
		int bufLen = 0;
		long pixelIdx = 0;

		for (int y = 0; y < height; y++)
		image.ProcessPixelRows(accessor =>
		{
		int read = reader.Read(rowBuf, 0, rowBuf.Length);
		if (read < rowBuf.Length)
		Array.Clear(rowBuf, read, rowBuf.Length - read);
		int curY = -1;
		Span<Rgba32> curRow = default;

		image.ProcessPixelRows(accessor =>
		while (pixelIdx < usedPixels)
		{
		var row = accessor.GetRowSpan(y);
		for (int x = 0; x < width; x++)
		// 确保缓冲区有至少2个采样（1帧）
		if (bufPos + 2 > bufLen)
		{
		long pixelIdx = (long)y * width + x;
		row[x] = DecodePixel(rowBuf, x * FramesPerPixel, audioChannels, pixelIdx);
		int remaining = bufLen - bufPos;
		if (remaining > 0)
		Array.Copy(sampleBuf, bufPos, sampleBuf, 0, remaining);
		bufPos = 0;
		bufLen = remaining + reader.Read(sampleBuf, remaining, sampleBuf.Length - remaining);
		if (bufLen < 2) break;
		}
		});

		decodedPixels += width;
		if (read < rowBuf.Length) break; // EOF：剩余行保持默认（透明）
		// 立体声帧：[left, right]
		float left = sampleBuf[bufPos];
		float right = sampleBuf[bufPos + 1];
		bufPos += 2;

		// 偶数像素 → right
		int y0 = (int)(pixelIdx / width);
		int x0 = (int)(pixelIdx % width);
		if (y0 != curY) { curY = y0; curRow = accessor.GetRowSpan(y0); }
		curRow[x0] = FloatToPixel(right);
		pixelIdx++;

		if (pixelIdx >= usedPixels) break;

		// 奇数像素 → left
		int y1 = (int)(pixelIdx / width);
		int x1 = (int)(pixelIdx % width);
		if (y1 != curY) { curY = y1; curRow = accessor.GetRowSpan(y1); }
		curRow[x1] = FloatToPixel(left);
		pixelIdx++;
		}
		});

		Save(image, opts.OutputPath, opts.Format);
		AnsiConsole.MarkupLine($"[green]已写入: {opts.OutputPath} ({opts.Format})[/]");
		@@ -153,19 +181,15 @@ public class AudioImage : IPlayground

		private static (int width, int height) ComputeDimensions(long totalPixels, int? reqWidth, int? reqHeight)
		{
		// 两者都设置：固定尺寸，多余采样丢弃
		if (reqWidth is { } w && reqHeight is { } h)
		return (w, h);

		// 只设置宽度：高度自动 = totalPixels / width
		if (reqWidth is { } w2)
		return (w2, (int)Math.Max(1, totalPixels / w2));

		// 只设置高度：宽度自动 = totalPixels / height
		if (reqHeight is { } h2)
		return ((int)Math.Max(1, totalPixels / h2), h2);

		// 都未设置：尽量正方形
		int side = (int)Math.Ceiling(Math.Sqrt(totalPixels));
		side = Math.Max(1, side);
		int autoHeight = (int)Math.Ceiling((double)totalPixels / side);
		@@ -173,32 +197,17 @@ public class AudioImage : IPlayground
		}

		/// <summary>
		/// 从 32 个音频帧还原 1 个 Rgba32 像素。
		/// 奇数像素取左声道，偶数像素取右声道（与 ImageAudio 正向映射对应）；
		/// 采样值 > 0 → bit 1，否则 → bit 0。
		/// float → RGBA 4字节（与 ImageAudio.PixelToFloat 互逆）
		/// </summary>
		private static Rgba32 DecodePixel(float[] buf, int frameBase, int audioChannels, long pixelIdx)
		{
		bool useLeft = (pixelIdx & 1) != 0;
		int channelOffset = useLeft ? 0 : (audioChannels >= 2 ? 1 : 0);

		byte r = 0, g = 0, b = 0, a = 0;
		for (int j = 0; j < FramesPerPixel; j++)
		{
		float v = buf[(frameBase + j) * audioChannels + channelOffset];
		if (v > 0f)
		{
		byte mask = (byte)(1 << (j & 7));
		switch (j >> 3)
		{
		case 0: r \|= mask; break;
		case 1: g \|= mask; break;
		case 2: b \|= mask; break;
		case 3: a \|= mask; break;
		}
		}
		}
		return new Rgba32(r, g, b, a);
		private static Rgba32 FloatToPixel(float v)
		{
		int bits = BitConverter.SingleToInt32Bits(v);
		return new Rgba32(
		(byte)(bits & 0xFF),
		(byte)((bits >> 8) & 0xFF),
		(byte)((bits >> 16) & 0xFF),
		(byte)((bits >> 24) & 0xFF)
		);
		}

		private static void Save(Image image, string path, string format)

TestV/Playgrounds/ImageAudio.cs

+54 −51

Original line number	Diff line number	Diff line
		@@ -12,9 +12,8 @@ public class ImageAudio : IPlayground
		{
		public string LaunchCmd { get; } = "ia";
		public string? Name { get; } = "Image Audio";
		public string? Description { get; } = "将32位像素转换成音频采样（奇数像素→左声道，偶数像素→右声道，支持WAV/MP3输出、Ctrl+C停止）";
		public string? Description { get; } = "将像素RGBA字节直接作为音频采样（奇数像素→左声道，偶数像素→右声道，可逆转换，支持WAV/MP3输出、Ctrl+C停止）";

		private const int BitsPerPixel = 32; // R/G/B/A 各 8 bit
		private const int Channels = 2; // 立体声
		private const int DefaultSampleRate = 44100;
		private const int DefaultMp3BitRate = 128; // kbps
		@@ -31,9 +30,13 @@ public class ImageAudio : IPlayground
		return;
		}

		if (!EnsureOutputAvailable(opts.WavPath)) return;
		if (!EnsureOutputAvailable(opts.Mp3Path)) return;

		using var image = Image.Load<Rgba32>(opts.ImagePath);
		int pixelCount = image.Width * image.Height;
		long totalFrames = (long)pixelCount * BitsPerPixel;
		// 每2个像素=1个立体声帧（偶数像素→右声道，奇数像素→左声道）
		long totalFrames = (pixelCount + 1) / 2;
		long totalSamples = totalFrames * Channels;
		long totalBytes = totalSamples * sizeof(float);

		@@ -74,6 +77,16 @@ public class ImageAudio : IPlayground
		}
		}

		private static bool EnsureOutputAvailable(string? path)
		{
		if (path is not null && File.Exists(path))
		{
		AnsiConsole.MarkupLine($"[red]输出文件已存在: {path}[/]");
		return false;
		}
		return true;
		}

		private sealed class Options
		{
		public required string ImagePath { get; init; }
		@@ -171,8 +184,8 @@ public class ImageAudio : IPlayground
		}

		/// <summary>
		/// 按需生成采样的 ISampleProvider。
		/// 奇数像素 → 左声道，偶数像素 → 右声道；每像素 32 个立体声帧。
		/// 可逆映射：像素的 RGBA 4字节直接作为一个 float 采样的4字节。
		/// 偶数像素 → 右声道，奇数像素 → 左声道；每2个像素产生1个立体声帧。
		/// 仅缓存一行像素，避免一次性持有全部采样。
		/// </summary>
		private sealed class PixelSampleProvider : ISampleProvider
		@@ -182,11 +195,7 @@ public class ImageAudio : IPlayground
		private readonly Rgba32[] _rowBuf;
		private int _rowY;
		private int _x;
		private int _bitInPixel;
		private long _pixelIdx;
		private Rgba32 _current;
		private bool _toLeft;
		private bool _colorWritten;

		public WaveFormat WaveFormat { get; }

		@@ -199,7 +208,6 @@ public class ImageAudio : IPlayground
		_rowY = 0;
		_x = 0;
		LoadRow(0);
		SetCurrent();
		}

		private void LoadRow(int y)
		@@ -207,19 +215,8 @@ public class ImageAudio : IPlayground
		_image.ProcessPixelRows(accessor => accessor.GetRowSpan(y).CopyTo(_rowBuf));
		}

		private void SetCurrent()
		private bool AdvancePixel()
		{
		_current = _rowBuf[_x];
		_toLeft = (_pixelIdx & 1) != 0;
		_colorWritten = false;
		_bitInPixel = 0;
		}

		private bool Advance()
		{
		_bitInPixel++;
		if (_bitInPixel < BitsPerPixel) return true;

		_pixelIdx++;
		_x++;
		if (_x >= _image.Width)
		@@ -229,55 +226,61 @@ public class ImageAudio : IPlayground
		if (_rowY >= _image.Height) return false;
		LoadRow(_rowY);
		}
		SetCurrent();
		return true;
		}

		public int Read(float[] buffer, int offset, int count)
		{
		int written = 0;
		int limit = count - (Channels - 1); // 保证能写满一个立体声帧
		while (written < limit)
		while (written + 1 < count)
		{
		if (_rowY >= _image.Height) break;

		if (_showColor && !_colorWritten)
		{
		AnsiConsole.Write(new Text(" ", new Style(null, new Color(_current.R, _current.G, _current.B))));
		_colorWritten = true;
		}
		// 偶数像素 → 右声道
		Rgba32 evenPixel = _rowBuf[_x];
		float right = PixelToFloat(evenPixel);

		if (_showColor)
		AnsiConsole.Write(new Text(" ", new Style(null, new Color(evenPixel.R, evenPixel.G, evenPixel.B))));

		float v = GetBitValue(_current, _bitInPixel);
		int idx = offset + written;
		if (_toLeft)
		if (!AdvancePixel())
		{
		buffer[idx] = v;
		buffer[idx + 1] = 0f;
		// 没有奇数像素了，只有右声道
		buffer[offset + written++] = 0f;
		buffer[offset + written++] = right;
		break;
		}
		else

		if (_rowY >= _image.Height)
		{
		buffer[idx] = 0f;
		buffer[idx + 1] = v;
		buffer[offset + written++] = 0f;
		buffer[offset + written++] = right;
		break;
		}
		written += Channels;

		if (!Advance()) break;
		// 奇数像素 → 左声道
		Rgba32 oddPixel = _rowBuf[_x];
		float left = PixelToFloat(oddPixel);

		if (_showColor)
		AnsiConsole.Write(new Text(" ", new Style(null, new Color(oddPixel.R, oddPixel.G, oddPixel.B))));

		// 立体声帧顺序：[left, right]
		buffer[offset + written++] = left;
		buffer[offset + written++] = right;

		if (!AdvancePixel()) break;
		}
		return written;
		}

		private static float GetBitValue(Rgba32 px, int bit)
		{
		// bit 0..7: R, 8..15: G, 16..23: B, 24..31: A
		// 二进制 0 → -1，二进制 1 → 1，使波形在 [-1, 1] 间波动
		byte channel = (bit >> 3) switch
		/// <summary>
		/// RGBA 4字节 → float（小端序，与 BitConverter.Int32BitsToSingle 对应）
		/// </summary>
		private static float PixelToFloat(Rgba32 px)
		{
		0 => px.R,
		1 => px.G,
		2 => px.B,
		_ => px.A,
		};
		return ((channel >> (bit & 7)) & 1) != 0 ? 1f : -1f;
		int bits = px.R \| (px.G << 8) \| (px.B << 16) \| (px.A << 24);
		return BitConverter.Int32BitsToSingle(bits);
		}
		}
		}

Admin message

Admin message