engine/codec/sps.go

package codec

import (
	"bytes"

	"github.com/deepch/vdk/codec/h265parser"
	"m7s.live/engine/v4/util/bits"
)

type SPSInfo struct {
	ProfileIdc uint
	LevelIdc   uint

	MbWidth  uint
	MbHeight uint

	CropLeft   uint
	CropRight  uint
	CropTop    uint
	CropBottom uint

	Width  uint
	Height uint
}

func ParseSPS(data []byte) (self SPSInfo, err error) {
	r := &bits.GolombBitReader{R: bytes.NewReader(data)}

	if _, err = r.ReadBits(8); err != nil {
		return
	}

	if self.ProfileIdc, err = r.ReadBits(8); err != nil {
		return
	}

	// constraint_set0_flag-constraint_set6_flag,reserved_zero_2bits
	if _, err = r.ReadBits(8); err != nil {
		return
	}

	// level_idc
	if self.LevelIdc, err = r.ReadBits(8); err != nil {
		return
	}

	// seq_parameter_set_id
	if _, err = r.ReadExponentialGolombCode(); err != nil {
		return
	}

	if self.ProfileIdc == 100 || self.ProfileIdc == 110 ||
		self.ProfileIdc == 122 || self.ProfileIdc == 244 ||
		self.ProfileIdc == 44 || self.ProfileIdc == 83 ||
		self.ProfileIdc == 86 || self.ProfileIdc == 118 {

		var chroma_format_idc uint
		if chroma_format_idc, err = r.ReadExponentialGolombCode(); err != nil {
			return
		}

		if chroma_format_idc == 3 {
			// residual_colour_transform_flag
			if _, err = r.ReadBit(); err != nil {
				return
			}
		}

		// bit_depth_luma_minus8
		if _, err = r.ReadExponentialGolombCode(); err != nil {
			return
		}
		// bit_depth_chroma_minus8
		if _, err = r.ReadExponentialGolombCode(); err != nil {
			return
		}
		// qpprime_y_zero_transform_bypass_flag
		if _, err = r.ReadBit(); err != nil {
			return
		}

		var seq_scaling_matrix_present_flag uint
		if seq_scaling_matrix_present_flag, err = r.ReadBit(); err != nil {
			return
		}

		if seq_scaling_matrix_present_flag != 0 {
			for i := 0; i < 8; i++ {
				var seq_scaling_list_present_flag uint
				if seq_scaling_list_present_flag, err = r.ReadBit(); err != nil {
					return
				}
				if seq_scaling_list_present_flag != 0 {
					var sizeOfScalingList uint
					if i < 6 {
						sizeOfScalingList = 16
					} else {
						sizeOfScalingList = 64
					}
					lastScale := uint(8)
					nextScale := uint(8)
					for j := uint(0); j < sizeOfScalingList; j++ {
						if nextScale != 0 {
							var delta_scale uint
							if delta_scale, err = r.ReadSE(); err != nil {
								return
							}
							nextScale = (lastScale + delta_scale + 256) % 256
						}
						if nextScale != 0 {
							lastScale = nextScale
						}
					}
				}
			}
		}
	}

	// log2_max_frame_num_minus4
	if _, err = r.ReadExponentialGolombCode(); err != nil {
		return
	}

	var pic_order_cnt_type uint
	if pic_order_cnt_type, err = r.ReadExponentialGolombCode(); err != nil {
		return
	}
	if pic_order_cnt_type == 0 {
		// log2_max_pic_order_cnt_lsb_minus4
		if _, err = r.ReadExponentialGolombCode(); err != nil {
			return
		}
	} else if pic_order_cnt_type == 1 {
		// delta_pic_order_always_zero_flag
		if _, err = r.ReadBit(); err != nil {
			return
		}
		// offset_for_non_ref_pic
		if _, err = r.ReadSE(); err != nil {
			return
		}
		// offset_for_top_to_bottom_field
		if _, err = r.ReadSE(); err != nil {
			return
		}
		var num_ref_frames_in_pic_order_cnt_cycle uint
		if num_ref_frames_in_pic_order_cnt_cycle, err = r.ReadExponentialGolombCode(); err != nil {
			return
		}
		for i := uint(0); i < num_ref_frames_in_pic_order_cnt_cycle; i++ {
			if _, err = r.ReadSE(); err != nil {
				return
			}
		}
	}

	// max_num_ref_frames
	if _, err = r.ReadExponentialGolombCode(); err != nil {
		return
	}

	// gaps_in_frame_num_value_allowed_flag
	if _, err = r.ReadBit(); err != nil {
		return
	}

	if self.MbWidth, err = r.ReadExponentialGolombCode(); err != nil {
		return
	}
	self.MbWidth++

	if self.MbHeight, err = r.ReadExponentialGolombCode(); err != nil {
		return
	}
	self.MbHeight++

	var frame_mbs_only_flag uint
	if frame_mbs_only_flag, err = r.ReadBit(); err != nil {
		return
	}
	if frame_mbs_only_flag == 0 {
		// mb_adaptive_frame_field_flag
		if _, err = r.ReadBit(); err != nil {
			return
		}
	}

	// direct_8x8_inference_flag
	if _, err = r.ReadBit(); err != nil {
		return
	}

	var frame_cropping_flag uint
	if frame_cropping_flag, err = r.ReadBit(); err != nil {
		return
	}
	if frame_cropping_flag != 0 {
		if self.CropLeft, err = r.ReadExponentialGolombCode(); err != nil {
			return
		}
		if self.CropRight, err = r.ReadExponentialGolombCode(); err != nil {
			return
		}
		if self.CropTop, err = r.ReadExponentialGolombCode(); err != nil {
			return
		}
		if self.CropBottom, err = r.ReadExponentialGolombCode(); err != nil {
			return
		}
	}

	self.Width = (self.MbWidth * 16) - self.CropLeft*2 - self.CropRight*2
	self.Height = ((2 - frame_mbs_only_flag) * self.MbHeight * 16) - self.CropTop*2 - self.CropBottom*2

	return
}

func ParseHevcSPS(data []byte) (self SPSInfo, err error) {
	var rawsps h265parser.SPSInfo
	if rawsps, err = h265parser.ParseSPS(data); err == nil {
		self.CropLeft, self.CropRight, self.CropTop, self.CropBottom = uint(rawsps.CropLeft), uint(rawsps.CropRight), uint(rawsps.CropTop), uint(rawsps.CropBottom)
		self.ProfileIdc, self.LevelIdc = rawsps.ProfileIdc, rawsps.LevelIdc
		self.MbWidth, self.MbHeight = rawsps.MbWidth, rawsps.MbHeight
		self.Width = uint(rawsps.Width)
		self.Height = uint(rawsps.Height)
	}
	return
}