Skip to content

Latest commit

 

History

History
140 lines (112 loc) · 7.36 KB

cambi.md

File metadata and controls

140 lines (112 loc) · 7.36 KB
Raw

CAMBI

CAMBI (Contrast Aware Multiscale Banding Index) is Netflix's detector for banding (aka contouring) artifacts.

Background

For an introduction to CAMBI, please refer to the tech blog. For a detailed technical description, please refer to the technical paper published at PCS 2021. Note that the paper describes an initial version of CAMBI that no longer matches the code exactly, but it is still a good introduction.

By default, the current version of CAMBI is a no-reference metric, and operates on a frame-by-frame basis (no temporal information is leveraged). To integrate it as part of the VMAF framework, which employs a full-reference metric API, CAMBI takes both a reference and a distorted video as its input. For simplicity, one can point the input arguments --reference and --distorted to the same video path.

CAMBI also offers a full-reference mode which computes its score as MAX(0, distorted_score - reference_score). This mode can be activated with the --full_ref command line option. In this case, both the --reference and --distorted inputs will be used.

Scores

The CAMBI score starts at 0, meaning no banding is detected. A higher CAMBI score means more visible banding artifacts are identified. The maximum CAMBI observed in a sequence is 24 (unwatchable). As a rule of thumb, a CAMBI score around 5 is where banding starts to become slightly annoying (also note that banding is highly dependent on the viewing environment - the brigher the display, and the dimmer the ambient light, the more visible banding is).

How to run CAMBI

To invoke CAMBI using the VMAF command line, follow the instruction and use cambi as the feature name. For example, after downloading the input video src01_hrc01_576x324.yuv, invoke CAMBI via:

libvmaf/build/tools/vmaf \
    --reference src01_hrc01_576x324.yuv \
    --distorted src01_hrc01_576x324.yuv \
    --width 576 --height 324 --pixel_format 420 --bitdepth 8 \
    --no_prediction --feature cambi --output /dev/stdout

This will yield the output:

<VMAF version="4b42f672">
  <params qualityWidth="576" qualityHeight="324" />
  <fyi fps="52.47" />
  <frames>
    <frame frameNum="0" cambi="0.848047" />
    <frame frameNum="1" cambi="0.723467" />
    ...
    <frame frameNum="46" cambi="0.994815" />
    <frame frameNum="47" cambi="1.019691" />
  </frames>
  <pooled_metrics>
    <metric name="cambi" min="0.509878" max="1.019691" mean="0.689250" harmonic_mean="0.681308" />
  </pooled_metrics>
  <aggregate_metrics />
</VMAF>

Bit depths

CAMBI supports the same input bit depths as VMAF: 8, 10, 12 and 16. However, the computations in CAMBI will always be performed at the 10-bit level, and the other formats will be converted to 10-bit as a preprocessing step.

Options

The CAMBI feature extractor also supports additional optional parameters as listed below:

  • window_size (min: 15, max: 127, default: 63): Window size to compute CAMBI (default: 63 corresponds to ~1 degree at 4K resolution and 1.5H)
  • topk (min: 0, max: 1.0, default: 0.6): Ratio of pixels for the spatial pooling computation
  • tvi_threshold (min: 0.0001, max: 1.0, default: 0.019): Visibilty threshold for luminance ΔL < tvi_threshold*L_mean for BT.1886
  • max_log_contrast (min: 0, max: 5, default: 2): Maximum contrast in log luma level (2^max_log_contrast) at 10-bits. Default 2 is equivalent to 4 luma levels at 10-bit and 1 luma level at 8-bit. The default is recommended for banding artifacts coming from video compression.
  • full_ref: optional flag (default: false) to run CAMBI as a full-reference metric, outputting the per-frame difference between the encoded and source images as well as the existing no-reference score.
  • enc_width and enc_height: Encoding/processing resolution to compute the banding score, useful in cases where scaling was applied to the input prior to the computation of metrics
  • src_width and src_height: Encoding/processing resolution to compute the banding score on the reference image, only used if full_ref=true.

An example using the enc_width and enc_height options on the input video KristenAndSara_1280x720_8bit_processed.yuv which has been encoded at 540p and later upscaled to 1280p (specifying the accurate encoding width and height as input allows CAMBI to more accurately assess the banding artifact):

libvmaf/build/tools/vmaf \
    --reference KristenAndSara_1280x720_8bit_processed.yuv \
    --distorted KristenAndSara_1280x720_8bit_processed.yuv \
    --width 1280 --height 720 --pixel_format 420 --bitdepth 8 \
    --no_prediction --feature cambi=enc_width=960:enc_height=540 --output /dev/stdout

This will yield the output:

<VMAF version="4b42f672">
  <params qualityWidth="1280" qualityHeight="720" />
  <fyi fps="40000.00" />
  <frames>
    <frame frameNum="0" cambi="1.218365" />
  </frames>
  <pooled_metrics>
    <metric name="cambi" min="1.218365" max="1.218365" mean="1.218365" harmonic_mean="1.218365" />
  </pooled_metrics>
  <aggregate_metrics />
</VMAF>

If no encoding width and height parameters are specified:

libvmaf/build/tools/vmaf \
    --reference KristenAndSara_1280x720_8bit_processed.yuv \
    --distorted KristenAndSara_1280x720_8bit_processed.yuv \
    --width 1280 --height 720 --pixel_format 420 --bitdepth 8 \
    --no_prediction --feature cambi --output /dev/stdout

The output will be:

<VMAF version="4b42f672">
  <params qualityWidth="1280" qualityHeight="720" />
  <fyi fps="47619.05" />
  <frames>
    <frame frameNum="0" cambi="0.341833" />
  </frames>
  <pooled_metrics>
    <metric name="cambi" min="0.341833" max="0.341833" mean="0.341833" harmonic_mean="0.341833" />
  </pooled_metrics>
  <aggregate_metrics />
</VMAF>

Python Library

CAMBI can also be invoked in the Python library. Use CambiFeatureExtractor as the feature extractor, and CambiQualityRunner as the quality runner. Use CambiFullReferenceFeatureExtractor and CambiFullReferenceQualityRunner to run the full-reference version of CAMBI.

    dis_path = VmafConfig.test_resource_path("yuv", "KristenAndSara_1280x720_8bit_processed.yuv")
    asset = Asset(dataset="test", content_id=0, asset_id=0,
                  workdir_root=VmafConfig.workdir_path(),
                  ref_path=dis_path,
                  dis_path=dis_path,
                  asset_dict={'width': 1280, 'height': 720,
                              'dis_enc_width': 960, 'dis_enc_height': 540})

        self.qrunner = CambiQualityRunner(
            [asset, asset_original],
            None, fifo_mode=False,
            result_store=None,
            optional_dict={}
        )
        self.qrunner.run(parallelize=True)
        results = self.qrunner.results

        # score: arithmetic mean score over all frames
        self.assertAlmostEqual(results[0]['Cambi_score'],
                               1.218365, places=4)