)]}'
{
  "commit": "52256cdbcaf07e637c964f92671dfc82321f2125",
  "tree": "fd770527e9d1dc1e5271cbf49e3d9d98d80be8e3",
  "parents": [
    "67b07c520d253955d1b4ea2d0fca2a7d3b905bba"
  ],
  "author": {
    "name": "Yunqing Wang",
    "email": "yunqingwang@google.com",
    "time": "Thu Jul 25 19:17:46 2013 -0700"
  },
  "committer": {
    "name": "Yunqing Wang",
    "email": "yunqingwang@google.com",
    "time": "Thu Jul 25 19:59:33 2013 -0700"
  },
  "message": "Modify static threshold calculation\n\nUsed 3 * standard_deviation in internal threshold calculation\ninstead of fit curve. This actually approached the algorithm\nbetter.\nFor comparison, similar tests were done:\nThe overall psnr loss is less than before.\n1. derf set:\nwhen static-thresh \u003d 1, psnr loss is 0.329%;\nwhen static-thresh \u003d 500, psnr loss is 0.970%;\n2. stdhd set:\nwhen static-thresh \u003d 1, psnr loss is 0.922%;\nwhen static-thresh \u003d 500, psnr loss is 1.307%;\n\nSimilar speedup is achieved. For example,\nclip            bitrate  static-thresh psnr    time\nakiyo(cif)       500        0          48.952  5.077s(50f)\nakiyo            500        500        48.866  4.169s(50f)\n\nparkjoy(1080p)   4000       0          30.388  78.20s(30f)\nparkjoy          4000       500        30.367  70.85s(30f)\n\nsunflower(1080p) 4000       0          44.402  74.55s(30f)\nsunflower        4000       500        44.414  68.69s(30f)\n\nChange-Id: Ic78833642ce1911dbbd1cb6c899a2d7e2dfcc1f3\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "5383c7fbac30b29dbb05e55fff16b026e592f783",
      "old_mode": 33188,
      "old_path": "vp9/encoder/vp9_rdopt.c",
      "new_id": "619f3e44ab056de1592d7e2151055d340d0f8e8d",
      "new_mode": 33188,
      "new_path": "vp9/encoder/vp9_rdopt.c"
    }
  ]
}