// Copyright 2022 Google LLC. All rights reserved.
// SPDX-License-Identifier: BSD-2-Clause

#include <algorithm>
#include <cmath>
#include <memory>
#include <tuple>

#include "avif/avif.h"
#include "aviftest_helpers.h"
#include "gtest/gtest.h"

using ::testing::Combine;
using ::testing::Values;
using ::testing::ValuesIn;

namespace libavif {
namespace {

//------------------------------------------------------------------------------

constexpr uint32_t kModifierSize = 4 * 4;

// Modifies the pixel values of a channel in image by modifier[] (row-ordered).
template <typename PixelType>
void ModifyImageChannel(avifRGBImage* image, uint32_t channel_offset,
                        const uint8_t modifier[kModifierSize]) {
  const uint32_t channel_count = avifRGBFormatChannelCount(image->format);
  assert(channel_offset < channel_count);
  for (uint32_t y = 0, i = 0; y < image->height; ++y) {
    PixelType* pixel =
        reinterpret_cast<PixelType*>(image->pixels + image->rowBytes * y);
    for (uint32_t x = 0; x < image->width; ++x, ++i) {
      pixel[channel_offset] += modifier[i % kModifierSize];
      pixel += channel_count;
    }
  }
}

void ModifyImageChannel(avifRGBImage* image, uint32_t channel_offset,
                        const uint8_t modifier[kModifierSize]) {
  if (image->depth <= 8) {
    ModifyImageChannel<uint8_t>(image, channel_offset, modifier);
  } else {
    ModifyImageChannel<uint16_t>(image, channel_offset, modifier);
  }
}

// Accumulates stats about the differences between the images a and b.
template <typename PixelType>
void GetDiffSumAndSqDiffSum(const avifRGBImage& a, const avifRGBImage& b,
                            int64_t* diff_sum, int64_t* abs_diff_sum,
                            int64_t* sq_diff_sum, int64_t* max_abs_diff) {
  const uint32_t channel_count = avifRGBFormatChannelCount(a.format);
  for (uint32_t y = 0; y < a.height; ++y) {
    const PixelType* row_a =
        reinterpret_cast<PixelType*>(a.pixels + a.rowBytes * y);
    const PixelType* row_b =
        reinterpret_cast<PixelType*>(b.pixels + b.rowBytes * y);
    for (uint32_t x = 0; x < a.width * channel_count; ++x) {
      const int64_t diff =
          static_cast<int64_t>(row_b[x]) - static_cast<int64_t>(row_a[x]);
      *diff_sum += diff;
      *abs_diff_sum += std::abs(diff);
      *sq_diff_sum += diff * diff;
      *max_abs_diff = std::max(*max_abs_diff, std::abs(diff));
    }
  }
}

void GetDiffSumAndSqDiffSum(const avifRGBImage& a, const avifRGBImage& b,
                            int64_t* diff_sum, int64_t* abs_diff_sum,
                            int64_t* sq_diff_sum, int64_t* max_abs_diff) {
  (a.depth <= 8) ? GetDiffSumAndSqDiffSum<uint8_t>(a, b, diff_sum, abs_diff_sum,
                                                   sq_diff_sum, max_abs_diff)
                 : GetDiffSumAndSqDiffSum<uint16_t>(
                       a, b, diff_sum, abs_diff_sum, sq_diff_sum, max_abs_diff);
}

// Returns the Peak Signal-to-Noise Ratio from accumulated stats.
double GetPsnr(double sq_diff_sum, double num_diffs, double max_abs_diff) {
  if (sq_diff_sum == 0.) {
    return 99.;  // Lossless.
  }
  const double distortion =
      sq_diff_sum / (num_diffs * max_abs_diff * max_abs_diff);
  return (distortion > 0.) ? std::min(-10 * std::log10(distortion), 98.9)
                           : 98.9;  // Not lossless.
}

//------------------------------------------------------------------------------

// To exercise the chroma subsampling loss, the input samples must differ in
// each of the RGB channels. Chroma subsampling expects the input RGB channels
// to be correlated to minimize the quality loss.
constexpr uint8_t kRedNoise[kModifierSize] = {
    7,  14, 11, 5,   // Random permutation of 16 values.
    4,  6,  8,  15,  //
    2,  9,  13, 3,   //
    12, 1,  10, 0};
constexpr uint8_t kGreenNoise[kModifierSize] = {
    3,  2,  12, 15,  // Random permutation of 16 values
    14, 10, 7,  13,  // that is somewhat close to kRedNoise.
    5,  1,  9,  0,   //
    8,  4,  11, 6};
constexpr uint8_t kBlueNoise[kModifierSize] = {
    0,  8,  14, 9,   // Random permutation of 16 values
    13, 12, 2,  7,   // that is somewhat close to kGreenNoise.
    3,  1,  11, 10,  //
    6,  15, 5,  4};

//------------------------------------------------------------------------------

class RGBToYUVTest
    : public testing::TestWithParam<
          std::tuple</*rgb_depth=*/int, /*yuv_depth=*/int, avifRGBFormat,
                     avifPixelFormat, avifRange, avifMatrixCoefficients,
                     /*add_noise=*/bool, /*rgb_step=*/uint32_t,
                     /*max_abs_average_diff=*/double, /*min_psnr=*/double>> {};

// Converts from RGB to YUV and back to RGB for all RGB combinations, separated
// by a color step for reasonable timing. If add_noise is true, also applies
// some noise to the input samples to exercise chroma subsampling.
TEST_P(RGBToYUVTest, ConvertWholeRange) {
  const int rgb_depth = std::get<0>(GetParam());
  const int yuv_depth = std::get<1>(GetParam());
  const avifRGBFormat rgb_format = std::get<2>(GetParam());
  const avifPixelFormat yuv_format = std::get<3>(GetParam());
  const avifRange yuv_range = std::get<4>(GetParam());
  const avifMatrixCoefficients matrix_coefficients = std::get<5>(GetParam());
  // Whether to add noise to the input RGB samples. Should only impact
  // subsampled chroma (4:2:2 and 4:2:0).
  const bool add_noise = std::get<6>(GetParam());
  // Testing each RGB combination would be more accurate but results are similar
  // with faster settings.
  const uint32_t rgb_step = std::get<7>(GetParam());
  // Thresholds to pass.
  const double max_abs_average_diff = std::get<8>(GetParam());
  const double min_psnr = std::get<9>(GetParam());
  // Deduced constants.
  const bool is_monochrome =
      (yuv_format == AVIF_PIXEL_FORMAT_YUV400);  // If so, only test grey input.
  const uint32_t rgb_max = (1 << rgb_depth) - 1;

  // The YUV upsampling treats the first and last rows and columns differently
  // than the remaining pairs of rows and columns. An image of 16 pixels is used
  // to test all these possibilities.
  static constexpr int width = 4;
  static constexpr int height = 4;
  std::unique_ptr<avifImage, decltype(&avifImageDestroy)> yuv(
      avifImageCreate(width, height, yuv_depth, yuv_format), avifImageDestroy);
  yuv->matrixCoefficients = matrix_coefficients;
  yuv->yuvRange = yuv_range;
  testutil::AvifRgbImage src_rgb(yuv.get(), rgb_depth, rgb_format);
  testutil::AvifRgbImage dst_rgb(yuv.get(), rgb_depth, rgb_format);
  const testutil::RgbChannelOffsets offsets =
      testutil::GetRgbChannelOffsets(rgb_format);

  // Alpha values are not tested here. Keep it opaque.
  if (avifRGBFormatHasAlpha(src_rgb.format)) {
    testutil::FillImageChannel(&src_rgb, offsets.a, rgb_max);
  }

  // Estimate the loss from converting RGB values to YUV and back.
  int64_t diff_sum = 0, abs_diff_sum = 0, sq_diff_sum = 0, max_abs_diff = 0;
  int64_t num_diffs = 0;
  const uint32_t max_value = rgb_max - (add_noise ? 15 : 0);
  for (uint32_t r = 0; r < max_value + rgb_step; r += rgb_step) {
    r = std::min(r, max_value);  // Test the maximum sample value even if it is
                                 // not a multiple of rgb_step.
    testutil::FillImageChannel(&src_rgb, offsets.r, r);
    if (add_noise) {
      ModifyImageChannel(&src_rgb, offsets.r, kRedNoise);
    }

    if (is_monochrome) {
      // Test only greyish input when converting to a single channel.
      testutil::FillImageChannel(&src_rgb, offsets.g, r);
      testutil::FillImageChannel(&src_rgb, offsets.b, r);
      if (add_noise) {
        ModifyImageChannel(&src_rgb, offsets.g, kGreenNoise);
        ModifyImageChannel(&src_rgb, offsets.b, kBlueNoise);
      }

      ASSERT_EQ(avifImageRGBToYUV(yuv.get(), &src_rgb, AVIF_RGB_TO_YUV_DEFAULT),
                AVIF_RESULT_OK);
      ASSERT_EQ(avifImageYUVToRGB(yuv.get(), &dst_rgb, AVIF_YUV_TO_RGB_DEFAULT),
                AVIF_RESULT_OK);
      GetDiffSumAndSqDiffSum(src_rgb, dst_rgb, &diff_sum, &abs_diff_sum,
                             &sq_diff_sum, &max_abs_diff);
      num_diffs += src_rgb.width * src_rgb.height * 3;  // Alpha is lossless.
    } else {
      for (uint32_t g = 0; g < max_value + rgb_step; g += rgb_step) {
        g = std::min(g, max_value);
        testutil::FillImageChannel(&src_rgb, offsets.g, g);
        if (add_noise) {
          ModifyImageChannel(&src_rgb, offsets.g, kGreenNoise);
        }
        for (uint32_t b = 0; b < max_value + rgb_step; b += rgb_step) {
          b = std::min(b, max_value);
          testutil::FillImageChannel(&src_rgb, offsets.b, b);
          if (add_noise) {
            ModifyImageChannel(&src_rgb, offsets.b, kBlueNoise);
          }

          ASSERT_EQ(
              avifImageRGBToYUV(yuv.get(), &src_rgb, AVIF_RGB_TO_YUV_DEFAULT),
              AVIF_RESULT_OK);
          ASSERT_EQ(
              avifImageYUVToRGB(yuv.get(), &dst_rgb, AVIF_YUV_TO_RGB_DEFAULT),
              AVIF_RESULT_OK);
          GetDiffSumAndSqDiffSum(src_rgb, dst_rgb, &diff_sum, &abs_diff_sum,
                                 &sq_diff_sum, &max_abs_diff);
          num_diffs +=
              src_rgb.width * src_rgb.height * 3;  // Alpha is lossless.
        }
      }
    }
  }

  // Stats and thresholds.
  // Note: The thresholds defined in this test are calibrated for libyuv fast
  //       paths. See reformat_libyuv.c. Slower non-libyuv conversions in
  //       libavif have a higher precision (using floating point operations).
  const double average_diff =
      static_cast<double>(diff_sum) / static_cast<double>(num_diffs);
  const double average_abs_diff =
      static_cast<double>(abs_diff_sum) / static_cast<double>(num_diffs);
  const double psnr = GetPsnr(static_cast<double>(sq_diff_sum),
                              static_cast<double>(num_diffs), rgb_max);
  EXPECT_LE(std::abs(average_diff), max_abs_average_diff);
  EXPECT_GE(psnr, min_psnr);

  // Print stats for convenience and easier threshold tuning.
  static constexpr const char* kAvifRgbFormatToString[] = {
      "RGB", "RGBA", "ARGB", "BGR", "BGRA", "ABGR"};
  std::cout << " RGB " << rgb_depth << " bits, YUV " << yuv_depth << " bits, "
            << kAvifRgbFormatToString[rgb_format] << ", "
            << avifPixelFormatToString(yuv_format) << ", "
            << (yuv_range ? "full" : "lmtd") << ", MC " << matrix_coefficients
            << ", " << (add_noise ? "noisy" : "plain") << ", avg "
            << average_diff << ", abs avg " << average_abs_diff << ", max "
            << max_abs_diff << ", PSNR " << psnr << "dB" << std::endl;
}

// Converts from RGB to YUV and back to RGB for multiple buffer dimensions to
// exercise stride computation and subsampling edge cases.
TEST_P(RGBToYUVTest, ConvertWholeBuffer) {
  const int rgb_depth = std::get<0>(GetParam());
  const int yuv_depth = std::get<1>(GetParam());
  const avifRGBFormat rgb_format = std::get<2>(GetParam());
  const avifPixelFormat yuv_format = std::get<3>(GetParam());
  const avifRange yuv_range = std::get<4>(GetParam());
  const avifMatrixCoefficients matrix_coefficients = std::get<5>(GetParam());
  // Whether to add noise to the input RGB samples.
  const bool add_noise = std::get<6>(GetParam());
  // Threshold to pass.
  const double min_psnr = std::get<9>(GetParam());
  // Deduced constants.
  const bool is_monochrome =
      (yuv_format == AVIF_PIXEL_FORMAT_YUV400);  // If so, only test grey input.
  const uint32_t rgb_max = (1 << rgb_depth) - 1;

  // Estimate the loss from converting RGB values to YUV and back.
  int64_t diff_sum = 0, abs_diff_sum = 0, sq_diff_sum = 0, max_abs_diff = 0;
  int64_t num_diffs = 0;
  for (int width : {1, 2, 127}) {
    for (int height : {1, 2, 251}) {
      std::unique_ptr<avifImage, decltype(&avifImageDestroy)> yuv(
          avifImageCreate(width, height, yuv_depth, yuv_format),
          avifImageDestroy);
      yuv->matrixCoefficients = matrix_coefficients;
      yuv->yuvRange = yuv_range;
      testutil::AvifRgbImage src_rgb(yuv.get(), rgb_depth, rgb_format);
      testutil::AvifRgbImage dst_rgb(yuv.get(), rgb_depth, rgb_format);
      const testutil::RgbChannelOffsets offsets =
          testutil::GetRgbChannelOffsets(rgb_format);

      // Fill the input buffer with whatever content.
      testutil::FillImageChannel(&src_rgb, offsets.r, /*value=*/0);
      testutil::FillImageChannel(&src_rgb, offsets.g, /*value=*/0);
      testutil::FillImageChannel(&src_rgb, offsets.b, /*value=*/0);
      if (add_noise) {
        ModifyImageChannel(&src_rgb, offsets.r, kRedNoise);
        ModifyImageChannel(&src_rgb, offsets.g,
                           is_monochrome ? kRedNoise : kGreenNoise);
        ModifyImageChannel(&src_rgb, offsets.b,
                           is_monochrome ? kRedNoise : kBlueNoise);
      }
      // Alpha values are not tested here. Keep it opaque.
      if (avifRGBFormatHasAlpha(src_rgb.format)) {
        testutil::FillImageChannel(&src_rgb, offsets.a, rgb_max);
      }

      ASSERT_EQ(avifImageRGBToYUV(yuv.get(), &src_rgb, AVIF_RGB_TO_YUV_DEFAULT),
                AVIF_RESULT_OK);
      ASSERT_EQ(avifImageYUVToRGB(yuv.get(), &dst_rgb, AVIF_YUV_TO_RGB_DEFAULT),
                AVIF_RESULT_OK);
      GetDiffSumAndSqDiffSum(src_rgb, dst_rgb, &diff_sum, &abs_diff_sum,
                             &sq_diff_sum, &max_abs_diff);
      num_diffs += src_rgb.width * src_rgb.height * 3;
    }
  }
  // max_abs_average_diff is not tested here because it is not meaningful for
  // only 3*3 conversions as it takes the maximum difference per conversion.
  // PSNR is averaged on all pixels so it can be tested here.
  EXPECT_GE(GetPsnr(static_cast<double>(sq_diff_sum),
                    static_cast<double>(num_diffs), rgb_max),
            min_psnr);
}

constexpr avifRGBFormat kAllRgbFormats[] = {
    AVIF_RGB_FORMAT_RGB, AVIF_RGB_FORMAT_RGBA, AVIF_RGB_FORMAT_ARGB,
    AVIF_RGB_FORMAT_BGR, AVIF_RGB_FORMAT_BGRA, AVIF_RGB_FORMAT_ABGR};

// avifMatrixCoefficients-typed constants for testing::Values() to work on MSVC.
constexpr avifMatrixCoefficients kMatrixCoefficientsBT601 =
    AVIF_MATRIX_COEFFICIENTS_BT601;
constexpr avifMatrixCoefficients kMatrixCoefficientsIdentity =
    AVIF_MATRIX_COEFFICIENTS_IDENTITY;

// This is the default avifenc setup when encoding from 8b PNG files to AVIF.
INSTANTIATE_TEST_SUITE_P(
    DefaultFormat, RGBToYUVTest,
    Combine(/*rgb_depth=*/Values(8),
            /*yuv_depth=*/Values(8), Values(AVIF_RGB_FORMAT_RGBA),
            Values(AVIF_PIXEL_FORMAT_YUV420), Values(AVIF_RANGE_FULL),
            Values(kMatrixCoefficientsBT601),
            /*add_noise=*/Values(true),
            /*rgb_step=*/Values(3),
            /*max_abs_average_diff=*/Values(0.1),  // The color drift is almost
                                                   // centered.
            /*min_psnr=*/Values(36.)  // Subsampling distortion is acceptable.
            ));

// Keeping RGB samples in full range and same or higher bit depth should not
// bring any loss in the roundtrip.
INSTANTIATE_TEST_SUITE_P(Identity8b, RGBToYUVTest,
                         Combine(/*rgb_depth=*/Values(8),
                                 /*yuv_depth=*/Values(8, 10, 12),
                                 ValuesIn(kAllRgbFormats),
                                 Values(AVIF_PIXEL_FORMAT_YUV444),
                                 Values(AVIF_RANGE_FULL),
                                 Values(kMatrixCoefficientsIdentity),
                                 /*add_noise=*/Values(true),
                                 /*rgb_step=*/Values(31),
                                 /*max_abs_average_diff=*/Values(0.),
                                 /*min_psnr=*/Values(99.)));
INSTANTIATE_TEST_SUITE_P(Identity10b, RGBToYUVTest,
                         Combine(/*rgb_depth=*/Values(10),
                                 /*yuv_depth=*/Values(10, 12),
                                 ValuesIn(kAllRgbFormats),
                                 Values(AVIF_PIXEL_FORMAT_YUV444),
                                 Values(AVIF_RANGE_FULL),
                                 Values(kMatrixCoefficientsIdentity),
                                 /*add_noise=*/Values(true),
                                 /*rgb_step=*/Values(101),
                                 /*max_abs_average_diff=*/Values(0.),
                                 /*min_psnr=*/Values(99.)));
INSTANTIATE_TEST_SUITE_P(Identity12b, RGBToYUVTest,
                         Combine(/*rgb_depth=*/Values(12),
                                 /*yuv_depth=*/Values(12),
                                 ValuesIn(kAllRgbFormats),
                                 Values(AVIF_PIXEL_FORMAT_YUV444),
                                 Values(AVIF_RANGE_FULL),
                                 Values(kMatrixCoefficientsIdentity),
                                 /*add_noise=*/Values(true),
                                 /*rgb_step=*/Values(401),
                                 /*max_abs_average_diff=*/Values(0.),
                                 /*min_psnr=*/Values(99.)));

// 4:4:4 and chroma subsampling have similar distortions on plain color inputs.
INSTANTIATE_TEST_SUITE_P(
    PlainAnySubsampling8b, RGBToYUVTest,
    Combine(
        /*rgb_depth=*/Values(8),
        /*yuv_depth=*/Values(8), ValuesIn(kAllRgbFormats),
        Values(AVIF_PIXEL_FORMAT_YUV444, AVIF_PIXEL_FORMAT_YUV422,
               AVIF_PIXEL_FORMAT_YUV420),
        Values(AVIF_RANGE_FULL), Values(kMatrixCoefficientsBT601),
        /*add_noise=*/Values(false),
        /*rgb_step=*/Values(17),
        /*max_abs_average_diff=*/Values(0.02),  // The color drift is centered.
        /*min_psnr=*/Values(49.)  // RGB>YUV>RGB distortion is barely
                                  // noticeable.
        ));

// Converting grey RGB samples to full-range monochrome of same or greater bit
// depth should be lossless.
INSTANTIATE_TEST_SUITE_P(MonochromeLossless8b, RGBToYUVTest,
                         Combine(/*rgb_depth=*/Values(8),
                                 /*yuv_depth=*/Values(8, 10, 12),
                                 ValuesIn(kAllRgbFormats),
                                 Values(AVIF_PIXEL_FORMAT_YUV400),
                                 Values(AVIF_RANGE_FULL),
                                 Values(kMatrixCoefficientsBT601),
                                 /*add_noise=*/Values(false),
                                 /*rgb_step=*/Values(1),
                                 /*max_abs_average_diff=*/Values(0.),
                                 /*min_psnr=*/Values(99.)));
INSTANTIATE_TEST_SUITE_P(MonochromeLossless10b, RGBToYUVTest,
                         Combine(/*rgb_depth=*/Values(10),
                                 /*yuv_depth=*/Values(10, 12),
                                 ValuesIn(kAllRgbFormats),
                                 Values(AVIF_PIXEL_FORMAT_YUV400),
                                 Values(AVIF_RANGE_FULL),
                                 Values(kMatrixCoefficientsBT601),
                                 /*add_noise=*/Values(false),
                                 /*rgb_step=*/Values(1),
                                 /*max_abs_average_diff=*/Values(0.),
                                 /*min_psnr=*/Values(99.)));
INSTANTIATE_TEST_SUITE_P(MonochromeLossless12b, RGBToYUVTest,
                         Combine(/*rgb_depth=*/Values(12),
                                 /*yuv_depth=*/Values(12),
                                 ValuesIn(kAllRgbFormats),
                                 Values(AVIF_PIXEL_FORMAT_YUV400),
                                 Values(AVIF_RANGE_FULL),
                                 Values(kMatrixCoefficientsBT601),
                                 /*add_noise=*/Values(false),
                                 /*rgb_step=*/Values(1),
                                 /*max_abs_average_diff=*/Values(0.),
                                 /*min_psnr=*/Values(99.)));

// Can be used to print the drift of all RGB to YUV conversion possibilities.
// Also used for coverage.
INSTANTIATE_TEST_SUITE_P(
    All8b, RGBToYUVTest,
    Combine(/*rgb_depth=*/Values(8),
            /*yuv_depth=*/Values(8, 10, 12), ValuesIn(kAllRgbFormats),
            Values(AVIF_PIXEL_FORMAT_YUV444, AVIF_PIXEL_FORMAT_YUV422,
                   AVIF_PIXEL_FORMAT_YUV420),
            Values(AVIF_RANGE_LIMITED, AVIF_RANGE_FULL),
            Values(kMatrixCoefficientsBT601),
            /*add_noise=*/Values(false, true),
            /*rgb_step=*/Values(61),  // High or it would be too slow.
            /*max_abs_average_diff=*/Values(1.),  // Not very accurate because
                                                  // of high rgb_step.
            /*min_psnr=*/Values(36.)));
INSTANTIATE_TEST_SUITE_P(
    All10b, RGBToYUVTest,
    Combine(/*rgb_depth=*/Values(10),
            /*yuv_depth=*/Values(8, 10, 12), ValuesIn(kAllRgbFormats),
            Values(AVIF_PIXEL_FORMAT_YUV444, AVIF_PIXEL_FORMAT_YUV422,
                   AVIF_PIXEL_FORMAT_YUV420),
            Values(AVIF_RANGE_LIMITED, AVIF_RANGE_FULL),
            Values(kMatrixCoefficientsBT601),
            /*add_noise=*/Values(false, true),
            /*rgb_step=*/Values(211),  // High or it would be too slow.
            /*max_abs_average_diff=*/Values(0.2),  // Not very accurate because
                                                   // of high rgb_step.
            /*min_psnr=*/Values(47.)));
INSTANTIATE_TEST_SUITE_P(
    All12b, RGBToYUVTest,
    Combine(/*rgb_depth=*/Values(12),
            /*yuv_depth=*/Values(8, 10, 12), ValuesIn(kAllRgbFormats),
            Values(AVIF_PIXEL_FORMAT_YUV444, AVIF_PIXEL_FORMAT_YUV422,
                   AVIF_PIXEL_FORMAT_YUV420),
            Values(AVIF_RANGE_LIMITED, AVIF_RANGE_FULL),
            Values(kMatrixCoefficientsBT601),
            /*add_noise=*/Values(false, true),
            /*rgb_step=*/Values(809),  // High or it would be too slow.
            /*max_abs_average_diff=*/Values(0.3),  // Not very accurate because
                                                   // of high rgb_step.
            /*min_psnr=*/Values(52.)));

// TODO: Test other matrix coefficients than identity and bt.601.

// This was used to estimate the quality loss of libyuv for RGB-to-YUV.
// Disabled because it takes a few minutes.
INSTANTIATE_TEST_SUITE_P(
    DISABLED_All8bTo8b, RGBToYUVTest,
    Combine(/*rgb_depth=*/Values(8),
            /*yuv_depth=*/Values(8), ValuesIn(kAllRgbFormats),
            Values(AVIF_PIXEL_FORMAT_YUV444, AVIF_PIXEL_FORMAT_YUV422,
                   AVIF_PIXEL_FORMAT_YUV420, AVIF_PIXEL_FORMAT_YUV400),
            Values(AVIF_RANGE_FULL, AVIF_RANGE_LIMITED),
            Values(kMatrixCoefficientsBT601),
            /*add_noise=*/Values(false, true),
            /*rgb_step=*/Values(3),  // way faster and 99% similar to rgb_step=1
            /*max_abs_average_diff=*/Values(10.),
            /*min_psnr=*/Values(10.)));

}  // namespace
}  // namespace libavif