示例#1
0
    void operator()(int size, const rs2::frame& rs_matrix, char* max_matrix)
    {
      const auto image = (const uint8_t *)(rs_matrix.get_data());
      auto matrix_out = (char*)(max_matrix);

      std::copy(image, image + size, matrix_out);
    }
    rs2::frame hole_filling_filter::prepare_target_frame(const rs2::frame& f, const rs2::frame_source& source)
    {
        // Allocate and copy the content of the input data to the target
        rs2::frame tgt = source.allocate_video_frame(_target_stream_profile, f, int(_bpp), int(_width), int(_height), int(_stride), _extension_type);

        memmove(const_cast<void*>(tgt.get_data()), f.get_data(), _current_frm_size_pixels * _bpp);
        return tgt;
    }
bool validate_ppf_results(rs2::frame origin_depth, rs2::frame result_depth, const ppf_test_config& reference_data, size_t frame_idx)
{
    std::vector<uint16_t> diff2orig;
    std::vector<uint16_t> diff2ref;

    // Basic sanity scenario with no filters applied.
    // validating domain transform in/out conversion. Requiring input=output
    bool domain_transform_only = (reference_data.downsample_scale == 1) &&
        (!reference_data.spatial_filter) && (!reference_data.temporal_filter);

    auto result_profile = result_depth.get_profile().as<rs2::video_stream_profile>();
    REQUIRE(result_profile);
    CAPTURE(result_profile.width());
    CAPTURE(result_profile.height());

    REQUIRE(result_profile.width() == reference_data.output_res_x);
    REQUIRE(result_profile.height() == reference_data.output_res_y);

    auto pixels = result_profile.width()*result_profile.height();
    diff2ref.resize(pixels);
    if (domain_transform_only)
        diff2orig.resize(pixels);

    // Pixel-by-pixel comparison of the resulted filtered depth vs data ercorded with external tool
    auto v1 = reinterpret_cast<const uint16_t*>(result_depth.get_data());
    auto v2 = reinterpret_cast<const uint16_t*>(reference_data._output_frames[frame_idx].data());

    for (auto i = 0; i < pixels; i++)
    {
        uint16_t diff = std::abs(*v1++ - *v2++);
        diff2ref[i] = diff;
    }

    // validating depth<->disparity domain transformation is lostless.
    if (domain_transform_only)
        REQUIRE(profile_diffs("./DomainTransform.txt",diff2orig, 0, 0, frame_idx));

    // Validate the filters
    // The differences between the reference code and librealsense implementation are byte-compared below
    return profile_diffs("./Filterstransform.txt", diff2ref, 0.f, 0, frame_idx);
}
示例#4
0
    rs2::frame yuy2rgb::process_frame(const rs2::frame_source& source, const rs2::frame& f)
    {
        auto p = f.get_profile();
        if (p.get() != _source_stream_profile.get())
        {
            _source_stream_profile = p;
            _target_stream_profile = p.clone(p.stream_type(), p.stream_index(), RS2_FORMAT_RGB8);
        }
        
        rs2::frame ret;

        auto vf = f.as<rs2::video_frame>();
        ret = source.allocate_video_frame(_target_stream_profile, f, _traget_bpp, 
            vf.get_width(), vf.get_height(), vf.get_width() * _traget_bpp, RS2_EXTENSION_VIDEO_FRAME);

        byte* planes[1];
        planes[0] = (byte*)ret.get_data();

        unpack_yuy2_rgb8(planes, (const byte*)f.get_data(), vf.get_width(), vf.get_height());

        return ret;
    }