shithub: libvpx

Download patch

ref: 797a2556eba26b2fbf89d339509ea9a8bb02dcc8
parent: 68340a34706b4db35bbb879dacff758c0a80bd1e
author: Alex Converse <aconverse@google.com>
date: Thu Jan 15 08:56:55 EST 2015 

Allow larger encoder configurations.

Allow changing colorspace in the encoder and increasing frame size.

Change-Id: I8e7c3b891af29ce420a15beb4f6f9c250245b2bb


--- a/test/encode_test_driver.h
+++ b/test/encode_test_driver.h
@@ -140,6 +140,12 @@
   }
 #endif
 
+  void Config(const vpx_codec_enc_cfg_t *cfg) {
+    const vpx_codec_err_t res = vpx_codec_enc_config_set(&encoder_, cfg);
+    ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
+    cfg_ = *cfg;
+  }
+
   void set_deadline(unsigned long deadline) {
     deadline_ = deadline;
   }
--- a/test/resize_test.cc
+++ b/test/resize_test.cc
@@ -261,9 +261,115 @@
   }
 }
 
+vpx_img_fmt_t CspForFrameNumber(int frame) {
+  if (frame < 10)
+    return VPX_IMG_FMT_I420;
+  if (frame < 20)
+    return VPX_IMG_FMT_I444;
+  return VPX_IMG_FMT_I420;
+}
+
+class ResizeCspTest : public ResizeTest {
+ protected:
+#if WRITE_COMPRESSED_STREAM
+  ResizeCspTest()
+      : ResizeTest(),
+        frame0_psnr_(0.0),
+        outfile_(NULL),
+        out_frames_(0) {}
+#else
+  ResizeCspTest() : ResizeTest(), frame0_psnr_(0.0) {}
+#endif
+
+  virtual ~ResizeCspTest() {}
+
+  virtual void BeginPassHook(unsigned int /*pass*/) {
+#if WRITE_COMPRESSED_STREAM
+    outfile_ = fopen("vp91-2-05-cspchape.ivf", "wb");
+#endif
+  }
+
+  virtual void EndPassHook() {
+#if WRITE_COMPRESSED_STREAM
+    if (outfile_) {
+      if (!fseek(outfile_, 0, SEEK_SET))
+        write_ivf_file_header(&cfg_, out_frames_, outfile_);
+      fclose(outfile_);
+      outfile_ = NULL;
+    }
+#endif
+  }
+
+  virtual void PreEncodeFrameHook(libvpx_test::VideoSource *video,
+                                  libvpx_test::Encoder *encoder) {
+    if (CspForFrameNumber(video->frame()) != VPX_IMG_FMT_I420 &&
+        cfg_.g_profile != 1) {
+      cfg_.g_profile = 1;
+      encoder->Config(&cfg_);
+    }
+    if (CspForFrameNumber(video->frame()) == VPX_IMG_FMT_I420 &&
+        cfg_.g_profile != 0) {
+      cfg_.g_profile = 0;
+      encoder->Config(&cfg_);
+    }
+  }
+
+  virtual void PSNRPktHook(const vpx_codec_cx_pkt_t *pkt) {
+    if (!frame0_psnr_)
+      frame0_psnr_ = pkt->data.psnr.psnr[0];
+    EXPECT_NEAR(pkt->data.psnr.psnr[0], frame0_psnr_, 2.0);
+  }
+
+#if WRITE_COMPRESSED_STREAM
+  virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {
+    ++out_frames_;
+
+    // Write initial file header if first frame.
+    if (pkt->data.frame.pts == 0)
+      write_ivf_file_header(&cfg_, 0, outfile_);
+
+    // Write frame header and data.
+    write_ivf_frame_header(pkt, outfile_);
+    (void)fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz, outfile_);
+  }
+#endif
+
+  double frame0_psnr_;
+#if WRITE_COMPRESSED_STREAM
+  FILE *outfile_;
+  unsigned int out_frames_;
+#endif
+};
+
+class ResizingCspVideoSource : public ::libvpx_test::DummyVideoSource {
+ public:
+  ResizingCspVideoSource() {
+    SetSize(kInitialWidth, kInitialHeight);
+    limit_ = 30;
+  }
+
+  virtual ~ResizingCspVideoSource() {}
+
+ protected:
+  virtual void Next() {
+    ++frame_;
+    SetImageFormat(CspForFrameNumber(frame_));
+    FillFrame();
+  }
+};
+
+TEST_P(ResizeCspTest, TestResizeCspWorks) {
+  ResizingCspVideoSource video;
+  cfg_.rc_min_quantizer = cfg_.rc_max_quantizer = 48;
+  cfg_.g_lag_in_frames = 0;
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+}
+
 VP8_INSTANTIATE_TEST_CASE(ResizeTest, ONE_PASS_TEST_MODES);
 VP9_INSTANTIATE_TEST_CASE(ResizeTest,
                           ::testing::Values(::libvpx_test::kRealTime));
 VP9_INSTANTIATE_TEST_CASE(ResizeInternalTest,
                           ::testing::Values(::libvpx_test::kOnePassBest));
+VP9_INSTANTIATE_TEST_CASE(ResizeCspTest,
+                          ::testing::Values(::libvpx_test::kRealTime));
 }  // namespace
--- a/test/video_source.h
+++ b/test/video_source.h
@@ -134,8 +134,13 @@
 
 class DummyVideoSource : public VideoSource {
  public:
-  DummyVideoSource() : img_(NULL), limit_(100), width_(0), height_(0) {
-    SetSize(80, 64);
+  DummyVideoSource()
+      : img_(NULL),
+        limit_(100),
+        width_(80),
+        height_(64),
+        format_(VPX_IMG_FMT_I420) {
+    ReallocImage();
   }
 
   virtual ~DummyVideoSource() { vpx_img_free(img_); }
@@ -174,17 +179,28 @@
 
   void SetSize(unsigned int width, unsigned int height) {
     if (width != width_ || height != height_) {
-      vpx_img_free(img_);
-      img_ = vpx_img_alloc(NULL, VPX_IMG_FMT_I420, width, height, 32);
-      raw_sz_ = ((img_->w + 31) & ~31) * img_->h * 3 / 2;
       width_ = width;
       height_ = height;
+      ReallocImage();
     }
   }
 
+  void SetImageFormat(vpx_img_fmt_t format) {
+    if (format_ != format) {
+      format_ = format;
+      ReallocImage();
+    }
+  }
+
  protected:
   virtual void FillFrame() { if (img_) memset(img_->img_data, 0, raw_sz_); }
 
+  void ReallocImage() {
+    vpx_img_free(img_);
+    img_ = vpx_img_alloc(NULL, format_, width_, height_, 32);
+    raw_sz_ = ((img_->w + 31) & ~31) * img_->h * img_->bps / 8;
+  }
+
   vpx_image_t *img_;
   size_t       raw_sz_;
   unsigned int limit_;
@@ -191,6 +207,7 @@
   unsigned int frame_;
   unsigned int width_;
   unsigned int height_;
+  vpx_img_fmt_t format_;
 };
 
 
--- a/vp9/encoder/vp9_encoder.c
+++ b/vp9/encoder/vp9_encoder.c
@@ -471,8 +471,9 @@
   VP9_COMMON *cm = &cpi->common;
   const VP9EncoderConfig *oxcf = &cpi->oxcf;
 
-  cpi->lookahead = vp9_lookahead_init(oxcf->width, oxcf->height,
-                                      cm->subsampling_x, cm->subsampling_y,
+  if (!cpi->lookahead)
+    cpi->lookahead = vp9_lookahead_init(oxcf->width, oxcf->height,
+                                        cm->subsampling_x, cm->subsampling_y,
 #if CONFIG_VP9_HIGHBITDEPTH
                                       cm->use_highbitdepth,
 #endif
@@ -1316,11 +1317,11 @@
   cm->height = cpi->oxcf.height;
 
   if (cpi->initial_width) {
-    // Increasing the size of the frame beyond the first seen frame, or some
-    // otherwise signaled maximum size, is not supported.
-    // TODO(jkoleszar): exit gracefully.
-    assert(cm->width <= cpi->initial_width);
-    assert(cm->height <= cpi->initial_height);
+    if (cm->width > cpi->initial_width || cm->height > cpi->initial_height) {
+      vp9_free_context_buffers(cm);
+      vp9_alloc_context_buffers(cm, cm->width, cm->height);
+      cpi->initial_width = cpi->initial_height = 0;
+    }
   }
   update_frame_size(cpi);
 
@@ -3439,7 +3440,11 @@
 #endif
   vpx_usec_timer_start(&timer);
 
-  if (vp9_lookahead_push(cpi->lookahead, sd, time_stamp, end_time, frame_flags))
+  if (vp9_lookahead_push(cpi->lookahead, sd, time_stamp, end_time,
+#if CONFIG_VP9_HIGHBITDEPTH
+                         use_highbitdepth,
+#endif  // CONFIG_VP9_HIGHBITDEPTH
+                         frame_flags))
     res = -1;
   vpx_usec_timer_mark(&timer);
   cpi->time_receive_data += vpx_usec_timer_elapsed(&timer);
--- a/vp9/encoder/vp9_lookahead.c
+++ b/vp9/encoder/vp9_lookahead.c
@@ -90,7 +90,11 @@
 #define USE_PARTIAL_COPY 0
 
 int vp9_lookahead_push(struct lookahead_ctx *ctx, YV12_BUFFER_CONFIG   *src,
-                       int64_t ts_start, int64_t ts_end, unsigned int flags) {
+                       int64_t ts_start, int64_t ts_end,
+#if CONFIG_VP9_HIGHBITDEPTH
+                       int use_highbitdepth,
+#endif
+                       unsigned int flags) {
   struct lookahead_entry *buf;
 #if USE_PARTIAL_COPY
   int row, col, active_end;
@@ -97,6 +101,13 @@
   int mb_rows = (src->y_height + 15) >> 4;
   int mb_cols = (src->y_width + 15) >> 4;
 #endif
+  int width = src->y_crop_width;
+  int height = src->y_crop_height;
+  int uv_width = src->uv_crop_width;
+  int uv_height = src->uv_crop_height;
+  int subsampling_x = src->subsampling_x;
+  int subsampling_y = src->subsampling_y;
+  int larger_dimensions, new_dimensions;
 
   if (ctx->sz + 1  + MAX_PRE_FRAMES > ctx->max_sz)
     return 1;
@@ -103,6 +114,16 @@
   ctx->sz++;
   buf = pop(ctx, &ctx->write_idx);
 
+  new_dimensions = width != buf->img.y_crop_width ||
+                   height != buf->img.y_crop_height ||
+                   uv_width != buf->img.uv_crop_width ||
+                   uv_height != buf->img.uv_crop_height;
+  larger_dimensions = width > buf->img.y_width ||
+                      height > buf->img.y_height ||
+                      uv_width > buf->img.uv_width ||
+                      uv_height > buf->img.uv_height;
+  assert(!larger_dimensions || new_dimensions);
+
 #if USE_PARTIAL_COPY
   // TODO(jkoleszar): This is disabled for now, as
   // vp9_copy_and_extend_frame_with_rect is not subsampling/alpha aware.
@@ -111,7 +132,7 @@
   // 1. Lookahead queue has has size of 1.
   // 2. Active map is provided.
   // 3. This is not a key frame, golden nor altref frame.
-  if (ctx->max_sz == 1 && active_map && !flags) {
+  if (!new_dimensions && ctx->max_sz == 1 && active_map && !flags) {
     for (row = 0; row < mb_rows; ++row) {
       col = 0;
 
@@ -147,11 +168,32 @@
       active_map += mb_cols;
     }
   } else {
+#endif
+    if (larger_dimensions) {
+      YV12_BUFFER_CONFIG new_img;
+      memset(&new_img, 0, sizeof(new_img));
+      if (vp9_alloc_frame_buffer(&new_img,
+                                 width, height, subsampling_x, subsampling_y,
+#if CONFIG_VP9_HIGHBITDEPTH
+                                 use_highbitdepth,
+#endif
+                                 VP9_ENC_BORDER_IN_PIXELS,
+                                 0))
+          return 1;
+      vp9_free_frame_buffer(&buf->img);
+      buf->img = new_img;
+    } else if (new_dimensions) {
+      buf->img.y_crop_width = src->y_crop_width;
+      buf->img.y_crop_height = src->y_crop_height;
+      buf->img.uv_crop_width = src->uv_crop_width;
+      buf->img.uv_crop_height = src->uv_crop_height;
+      buf->img.subsampling_x = src->subsampling_x;
+      buf->img.subsampling_y = src->subsampling_y;
+    }
+    // Partial copy not implemented yet
     vp9_copy_and_extend_frame(src, &buf->img);
+#if USE_PARTIAL_COPY
   }
-#else
-  // Partial copy not implemented yet
-  vp9_copy_and_extend_frame(src, &buf->img);
 #endif
 
   buf->ts_start = ts_start;
--- a/vp9/encoder/vp9_lookahead.h
+++ b/vp9/encoder/vp9_lookahead.h
@@ -79,7 +79,11 @@
  * \param[in] active_map  Map that specifies which macroblock is active
  */
 int vp9_lookahead_push(struct lookahead_ctx *ctx, YV12_BUFFER_CONFIG *src,
-                       int64_t ts_start, int64_t ts_end, unsigned int flags);
+                       int64_t ts_start, int64_t ts_end,
+#if CONFIG_VP9_HIGHBITDEPTH
+                       int use_highbitdepth,
+#endif
+                       unsigned int flags);
 
 
 /**\brief Get the next source buffer to encode
--- a/vp9/vp9_cx_iface.c
+++ b/vp9/vp9_cx_iface.c
@@ -530,10 +530,9 @@
   if (cfg->g_w != ctx->cfg.g_w || cfg->g_h != ctx->cfg.g_h) {
     if (cfg->g_lag_in_frames > 1 || cfg->g_pass != VPX_RC_ONE_PASS)
       ERROR("Cannot change width or height after initialization");
-    if ((ctx->cpi->initial_width && (int)cfg->g_w > ctx->cpi->initial_width) ||
+    if (!valid_ref_frame_size(ctx->cfg.g_w, ctx->cfg.g_h, cfg->g_w, cfg->g_h) ||
+        (ctx->cpi->initial_width && (int)cfg->g_w > ctx->cpi->initial_width) ||
         (ctx->cpi->initial_height && (int)cfg->g_h > ctx->cpi->initial_height))
-      ERROR("Cannot increase width or height larger than their initial values");
-    if (!valid_ref_frame_size(ctx->cfg.g_w, ctx->cfg.g_h, cfg->g_w, cfg->g_h))
       force_key = 1;
   }