Is there another way to export a frame in ffmpeg t

问题:

Sound is working in Linux the same as it did in Windows. But the video is just a black screen and when I attempt to save the frames as BMP files all of them were corrupt/empty files. I am using Ffmpeg.Autogen to interface with the libraries. https://github.com/Ruslan-B/FFmpeg.AutoGen. The file is VP8 and OGG in a MKV container. Though the extension is AVI for some reason.

I tried messing with the order of the code a bit. I checked to make sure the build of Ffmpeg on Linux had VP8. I was searching online but was having trouble finding another way to do what I am doing. This is to contribute to the OpenVIII project. My fork-> https://github.com/Sebanisu/OpenVIII

This just preps the scaler to change the pixelformat or else people have blue faces.

        private void PrepareScaler()
        {

            if (MediaType != AVMediaType.AVMEDIA_TYPE_VIDEO)
            {
                return;
            }

            ScalerContext = ffmpeg.sws_getContext(
                Decoder.CodecContext->width, Decoder.CodecContext->height, Decoder.CodecContext->pix_fmt,
                Decoder.CodecContext->width, Decoder.CodecContext->height, AVPixelFormat.AV_PIX_FMT_RGBA,
                ffmpeg.SWS_ACCURATE_RND, null, null, null);
            Return = ffmpeg.sws_init_context(ScalerContext, null, null);

            CheckReturn();
        }

Converts Frame to BMP I am thinking this is where the problem is. Because I had added bitmap.save to this and got empty BMPs.

public Bitmap FrameToBMP()
        {
            Bitmap bitmap = null;
            BitmapData bitmapData = null;

            try
            {
                bitmap = new Bitmap(Decoder.CodecContext->width, Decoder.CodecContext->height, PixelFormat.Format32bppArgb);
                AVPixelFormat v = Decoder.CodecContext->pix_fmt;

                // lock the bitmap
                bitmapData = bitmap.LockBits(new Rectangle(0, 0, Decoder.CodecContext->width, Decoder.CodecContext->height), ImageLockMode.ReadOnly, PixelFormat.Format32bppArgb);

                byte* ptr = (byte*)(bitmapData.Scan0);

                byte*[] srcData = { ptr, null, null, null };
                int[] srcLinesize = { bitmapData.Stride, 0, 0, 0 };

                // convert video frame to the RGB bitmap
                ffmpeg.sws_scale(ScalerContext, Decoder.Frame->data, Decoder.Frame->linesize, 0, Decoder.CodecContext->height, srcData, srcLinesize); //sws_scale broken on linux?
            }
            finally
            {
                if (bitmap != null && bitmapData != null)
                {
                    bitmap.UnlockBits(bitmapData);
                }
            }
            return bitmap;

        }

After I get a bitmap we turn it into a Texture2D so we can draw it.

 public Texture2D FrameToTexture2D()
        {
            //Get Bitmap. there might be a way to skip this step.
            using (Bitmap frame = FrameToBMP())
            {
                //string filename = Path.Combine(Path.GetTempPath(), $"{Path.GetFileNameWithoutExtension(DecodedFileName)}_rawframe.{Decoder.CodecContext->frame_number}.bmp");

                //frame.Save(filename);
                BitmapData bmpdata = null;
                Texture2D frameTex = null;
                try
                {
                    //Create Texture
                    frameTex = new Texture2D(Memory.spriteBatch.GraphicsDevice, frame.Width, frame.Height, false, SurfaceFormat.Color); //GC will collect frameTex
                                                                                                                                        //Fill it with the bitmap.
                    bmpdata = frame.LockBits(new Rectangle(0, 0, frame.Width, frame.Height), System.Drawing.Imaging.ImageLockMode.ReadOnly, PixelFormat.Format32bppArgb);// System.Drawing.Imaging.PixelFormat.Format32bppArgb);
                    byte[] texBuffer = new byte[bmpdata.Width * bmpdata.Height * 4]; //GC here
                    Marshal.Copy(bmpdata.Scan0, texBuffer, 0, texBuffer.Length);

                    frameTex.SetData(texBuffer);


                }
                finally
                {
                    if (bmpdata != null)
                    {
                        frame.UnlockBits(bmpdata);
                    }
                }
                return frameTex;

            }
        }

I can post more if you want it's pretty much all up on my fork

Video will play back as it does in Windows. As smooth as 15 fps can be. :)

回答1:

I ended up removing the bitmap part of the code. And it worked! So previously I would convert the frame to a bitmap and it would copy the pixels out of the bitmap into the Texture2D. I looked closer and realized I could skip that step of the bitmap. I'm sorry for not being clear enough on my question.

        /// <summary>
        /// Converts Frame to Texture
        /// </summary>
        /// <returns>Texture2D</returns>
        public Texture2D FrameToTexture2D()
        {
            Texture2D frameTex = new Texture2D(Memory.spriteBatch.GraphicsDevice, Decoder.CodecContext->width, Decoder.CodecContext->height, false, SurfaceFormat.Color);
            const int bpp = 4;
            byte[] texBuffer = new byte[Decoder.CodecContext->width * Decoder.CodecContext->height * bpp]; 
            fixed (byte* ptr = &texBuffer[0])
            {
                byte*[] srcData = { ptr, null, null, null };
                int[] srcLinesize = { Decoder.CodecContext->width * bpp, 0, 0, 0 };
                // convert video frame to the RGB data
                ffmpeg.sws_scale(ScalerContext, Decoder.Frame->data, Decoder.Frame->linesize, 0, Decoder.CodecContext->height, srcData, srcLinesize);
            }
            frameTex.SetData(texBuffer);
            return frameTex;
        }