First you need to create a SwsContext
(you need to do this only once) :
struct SwsContext *resize;
resize = sws_getContext(width1, height1, AV_PIX_FMT_YUV420P, width2, height2, PIX_FMT_RGB24, SWS_BICUBIC, NULL, NULL, NULL);
You need two frames for conversion, frame1 is the original frame, you need to explicitly allocate frame2 :
AVFrame* frame1 = avcodec_alloc_frame(); // this is your original frame
AVFrame* frame2 = avcodec_alloc_frame();
int num_bytes = avpicture_get_size(AV_PIX_FMT_RGB24, width2, height2);
uint8_t* frame2_buffer = (uint8_t *)av_malloc(num_bytes*sizeof(uint8_t));
avpicture_fill((AVPicture*)frame2, frame2_buffer, AV_PIX_FMT_RGB24, width2, height2);
You may use this part inside a loop if you need to resize each frame you receive :
// frame1 should be filled by now (eg using avcodec_decode_video)
sws_scale(resize, frame1->data, frame1->linesize, 0, height1, frame2->data, frame2->linesize);
Note that I also changed pixel format, but you can use the same pixel format for both frames
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…