如何将FFmpeg视频帧转换为YUV444？SDL播放器画质升级问询

Hey there! Let's walk through exactly how to convert FFmpeg video frames to YUV444 format for your SDL-based video player, while also addressing those deprecated function headaches from your outdated tutorial.

Core Concept: Using FFmpeg's Swscale Library #

The key tool here is FFmpeg's libswscale, which handles pixel format conversion (and scaling, if you need it). We'll be converting your input frame to AV_PIX_FMT_YUV444P (planar YUV444, where each color component is stored in a separate plane—perfect for SDL handling if you set things up right).

Step-by-Step Implementation #

1. Initialize the Swscale Context

First, create a conversion context that tells FFmpeg how to map your source frame to YUV444. Replace width, height, and frame_src->format with your actual input frame's properties:

#include <libswscale/swscale.h>

struct SwsContext *sws_ctx = sws_getContext(
    width, height, frame_src->format,  // Source specs
    width, height, AV_PIX_FMT_YUV444P, // Target specs (YUV444 planar)
    SWS_BILINEAR, NULL, NULL, NULL     // Scaling algorithm (bilinear is smooth)
);

if (!sws_ctx) {
    fprintf(stderr, "Failed to create Swscale context\n");
    // Add your error handling here
}

2. Allocate a Target Frame for YUV444

Next, create an empty frame to hold the converted YUV444 data, and allocate buffer memory for it:

AVFrame *frame_yuv444 = av_frame_alloc();
if (!frame_yuv444) {
    fprintf(stderr, "Failed to allocate YUV444 frame\n");
    sws_freeContext(sws_ctx);
    // Error handling
}

// Set frame properties to match YUV444P
frame_yuv444->format = AV_PIX_FMT_YUV444P;
frame_yuv444->width = width;
frame_yuv444->height = height;

// Allocate buffer memory for the frame
if (av_frame_get_buffer(frame_yuv444, 0) < 0) {
    fprintf(stderr, "Failed to allocate buffer for YUV444 frame\n");
    av_frame_free(&frame_yuv444);
    sws_freeContext(sws_ctx);
    // Error handling
}

3. Perform the Format Conversion

Now run the conversion to turn your source frame into YUV444:

sws_scale(
    sws_ctx,
    (const uint8_t *const *)frame_src->data, // Source data planes
    frame_src->linesize,                    // Source line sizes
    0, height,                              // Source region (full frame)
    frame_yuv444->data,                     // Target data planes
    frame_yuv444->linesize                  // Target line sizes
);

4. Render YUV444 with SDL

SDL doesn't have a built-in YUV444 texture format, but you have two solid options:

Option A: Render Separate Y/U/V Planes with Shaders (Best for True YUV444)

Create three grayscale textures for Y, U, and V planes, then use an SDL shader to combine them into RGB for display:

// Create Y/U/V textures (all 8-bit grayscale)
SDL_Texture *tex_y = SDL_CreateTexture(renderer, SDL_PIXELFORMAT_Y8, SDL_TEXTUREACCESS_STREAMING, width, height);
SDL_Texture *tex_u = SDL_CreateTexture(renderer, SDL_PIXELFORMAT_Y8, SDL_TEXTUREACCESS_STREAMING, width, height);
SDL_Texture *tex_v = SDL_CreateTexture(renderer, SDL_PIXELFORMAT_Y8, SDL_TEXTUREACCESS_STREAMING, width, height);

// Update each texture with the corresponding plane data
SDL_UpdateTexture(tex_y, NULL, frame_yuv444->data[0], frame_yuv444->linesize[0]);
SDL_UpdateTexture(tex_u, NULL, frame_yuv444->data[1], frame_yuv444->linesize[1]);
SDL_UpdateTexture(tex_v, NULL, frame_yuv444->data[2], frame_yuv444->linesize[2]);

// Use an SDL shader to merge YUV444 to RGB (you'll need a simple GLSL shader here)
// Example shader logic: Convert YUV444 to RGB using standard colorspace formulas

Option B: Convert YUV444 to RGB for Easy SDL Rendering

If you don't mind converting to RGB (you still get the high quality of YUV444 before conversion), adjust your Swscale context to target RGB32 instead:

// Change the target format in sws_getContext to AV_PIX_FMT_RGB32
struct SwsContext *sws_ctx = sws_getContext(
    width, height, frame_src->format,
    width, height, AV_PIX_FMT_RGB32,
    SWS_BILINEAR, NULL, NULL, NULL
);

// Then create an SDL RGB32 texture and update it directly
SDL_Texture *rgb_tex = SDL_CreateTexture(renderer, SDL_PIXELFORMAT_RGB32, SDL_TEXTUREACCESS_STREAMING, width, height);
SDL_UpdateTexture(rgb_tex, NULL, frame_rgb->data[0], frame_rgb->linesize[0]);

5. Clean Up Resources

Don't forget to free memory to avoid leaks:

av_frame_free(&frame_yuv444);
sws_freeContext(sws_ctx);
// Destroy SDL textures when done
SDL_DestroyTexture(tex_y);
SDL_DestroyTexture(tex_u);
SDL_DestroyTexture(tex_v);
// Or SDL_DestroyTexture(rgb_tex) if using Option B

Fixing Deprecated FFmpeg Functions #

Since your tutorial is outdated, here's a quick fix for common deprecated calls:

• Replace avcodec_decode_video2 with avcodec_send_packet + avcodec_receive_frame:

  if (avcodec_send_packet(codec_ctx, pkt) == 0) {
      while (avcodec_receive_frame(codec_ctx, frame_src) == 0) {
          // Process the frame (convert to YUV444 here)
      }
  }
  

• Ensure you're using avformat_open_input (still valid) with modern error checking, and avcodec_find_decoder instead of older decoder lookup functions.

内容的提问来源于stack exchange，提问作者Edward Severinsen