azayaka package

Submodules

azayaka.fileformat module

Azayaka:

SAR/InSAR Format Module. - JAXA CEOS Format

  • ALOS PALSAR

    • L1.0 RAW, L1.1 SLC

  • ALOS-2 PALSAR-2

    • L1.1 SLC

  • ALOS-4 PALSAR-3

    • L1.2 RAW, L1.1 SLC

  • ESA SAFE Format (TODO 3)

    • Sentinel-1

  • NGA Fromat (TODO 1)

    • SICD CPHD

  • NASA/ISRO CEOS Format (TODO 2)
    • NISAR

      • RAW, SLC

Copyright (c) 2026 Syusuke Yasui, Yutaka Yamamoto, and contributors. Licensed under the APGL-3.0 License.

Equation - Condition: No 2.xxx File.

class azayaka.fileformat.CEOS_PALSAR2_L11_SLC(PATH_CEOS_FOLDER: str, POLARIMETORY: str = 'HH', ORBIT_NAME: str = 'A')[source]

Bases: object

ALOS-2 PALSAR-2 Level 1.1 CEOS Format Reader

  • IMG: SAR image file (single-look complex, slant range, REAL*4 I/Q)

  • LED: SAR leader file (orbit, attitude, radiometric info)

BYTE1 = 1
BYTE2 = 2
DIGIT4 = 1000.0
FLOAT16 = 16
FLOAT22 = 22
FLOAT32 = 32
FLOAT64 = 64
INTERGER4 = 4
INTERGER6 = 6
NUM_VELOCITY_CALC_SPAN_COUNT: int = 4
SOL = 299792458.0
TIME_DAY_HOUR = 24
TIME_DAY_MINITE = 60
TIME_DAY_SEC = 86400
TIME_MINITE_SEC = 60
__init__(PATH_CEOS_FOLDER: str, POLARIMETORY: str = 'HH', ORBIT_NAME: str = 'A')[source]

Initialize CEOS Format Reader (ALOS-2 PALSAR-2 Level 1.1)

Parameters:

  • PATH_CEOS_FOLDER (str) – CEOS プロダクトが展開されているフォルダ 例: ALOS2267860740-150210/

  • POLARIMETORY (str, optional) – 偏波名 ‘HH’, ‘HV’, ‘VV’, ‘VH’.

  • ORBIT_NAME (str, optional) – 軌道名 ‘A’ or ‘D’.

set_geometory(plot: bool = False, PATH_OUTPUT: str | None = None, output_json_path: str = None)[source]

観測ジオメトリの設定

IMG/LED から読み出した軌道情報を用いて、観測中の 衛星位置ベクトル・速度ベクトル・スラントレンジサンプル などを計算する（ALOS-2 PALSAR-2 仕様に対応）。

class azayaka.fileformat.CEOS_PALSAR3_L11_SLC(PATH_CEOS_FOLDER: str, POLARIMETORY: str = 'HH', ORBIT_NAME: str = 'A')[source]

Bases: object

ALOS-4 PALSAR-3 Level 1.1 CEOS Format Reader

  • IMG: SAR image file (single-look complex, slant range, REAL*4 I/Q)

  • LED: SAR leader file (orbit, attitude, radiometric info)

ALOS-2 PALSAR-2 の CEOS フォーマットを踏襲した ALOS-4 PALSAR-3 標準プロダクトフォーマット (FTR-240031A) に対応。

BYTE1 = 1
BYTE2 = 2
DIGIT4 = 1000.0
FLOAT16 = 16
FLOAT22 = 22
FLOAT32 = 32
FLOAT64 = 64
INTERGER4 = 4
INTERGER6 = 6
NUM_VELOCITY_CALC_SPAN_COUNT: int = 4
SOL = 299792458.0
TIME_DAY_HOUR = 24
TIME_DAY_MINITE = 60
TIME_DAY_SEC = 86400
TIME_MINITE_SEC = 60
__init__(PATH_CEOS_FOLDER: str, POLARIMETORY: str = 'HH', ORBIT_NAME: str = 'A')[source]

Initialize CEOS Format Reader (ALOS-4 PALSAR-3 Level 1.1)

Parameters:

  • PATH_CEOS_FOLDER (str) – CEOS プロダクトが展開されているフォルダ 例: ALOS4xxxxxx…/

  • POLARIMETORY (str, optional) – 偏波名 ‘HH’, ‘HV’, ‘VV’, ‘VH’.

  • ORBIT_NAME (str, optional) – 軌道名 ‘A’ or ‘D’（ALOS-4 では ProductID 内に 含まれるが、互換のためパラメータとして残す）。

set_geometory(plot: bool = False, PATH_OUTPUT: str | None = None, output_json_path: str = None)[source]

観測ジオメトリの設定

IMG/LED から読み出した軌道情報を用いて、観測中の 衛星位置ベクトル・速度ベクトル・スラントレンジサンプル などを計算する（ALOS-4 PALSAR-3 仕様に対応）。

class azayaka.fileformat.CEOS_PALSAR_L10_RAW(PATH_CEOS_FOLDER: str, POLARIMETORY: str = 'HH', ORBIT_NAME: str = 'A')[source]

Bases: object

CEOS PALSAR Level 1.0 RAW reader.

Parses JAXA CEOS records and exposes acquisition metadata and raw signal data.

BYTE1 = 1
BYTE2 = 2
DIGIT4 = 1000.0
FLOAT16 = 16
FLOAT22 = 22
FLOAT32 = 32
FLOAT64 = 64
INTERGER4 = 4
INTERGER6 = 6
NUM_TMP_SAMPLE: int = 36
NUM_VELOCITY_CALC_SAMPLE: int = 4
NUM_VELOCITY_CALC_SPAN_COUNT: int = 4
SOL = 299792458.0
TIME_DAY_HOUR = 24
TIME_DAY_MINITE = 60
TIME_DAY_SEC = 86400
TIME_MINITE_SEC = 60
classmethod __init__(PATH_CEOS_FOLDER: str, POLARIMETORY: str = 'HH', ORBIT_NAME: str = 'A')[source]

Initialize CEOS Format Reader :param PATH_CEOS_FOLDER: Path to CEOS folder :type PATH_CEOS_FOLDER: str :param POLARIMETORY: Polarimetry mode. Defaults to ‘HH’, ‘HV’, ‘VV’, ‘VH’. :type POLARIMETORY: str, optional :param ORBIT_NAME: Orbit name, ‘A’ or ‘D’. Defaults to ‘A’. :type ORBIT_NAME: str, optional

execute_focus(ground_velocity=None, PATH_OUTPUT: str = './output')[source]

フォーカシングの実行（チャープスケーリング法）。

SAR の生データに対してチャープスケーリング（Chirp Scaling, CS）アルゴリズムを適用し、 レンジ方向とアジマス方向の補正・圧縮を行って複素画像を生成します。

処理の流れ（概要）: 1. 観測ジオメトリを準備し、衛星位置・速度から中心時刻の速度を推定します

（ground_velocity が指定されている場合はその値で上書き）。

  1. レンジ時間軸 TAU とドップラ中心 F_DOPPLER_CENTROID を定義します。

  2. H1: チャープスケーリングによるレンジ方向の位相補正。

  3. H2: バルク RCMC とレンジ圧縮。

  4. H3: 角度補正（位相誤差補正）。

  5. H4: アジマス圧縮。

  6. FFT/IFFT を用いて周波数領域と時刻領域を往復し、最終的な複素画像を返します。

Parameters:

  • ground_velocity (Optional[float]) – 地上速度の手動指定。None の場合は推定値を使用します。

  • PATH_OUTPUT (str) – ジオメトリ生成時の出力先フォルダ。

Returns:

フォーカシング後の複素画像 (azimuth x range)。

Return type:

np.ndarray

set_geometory(plot=False, output_json_path: str = None, PATH_OUTPUT: str = './output')[source]

観測ジオメトリの設定

class azayaka.fileformat.CEOS_PALSAR_L11_SLC(PATH_CEOS_FOLDER: str, POLARIMETORY: str = 'HH', ORBIT_NAME: str = 'A')[source]

Bases: object

CEOS PALSAR Level 1.1 SLC reader.

Parses ALOS PALSAR Level 1.1 CEOS products and exposes SLC data and metadata.

BYTE1 = 1
BYTE2 = 2
DIGIT4 = 1000.0
FLOAT16 = 16
FLOAT22 = 22
FLOAT32 = 32
FLOAT64 = 64
INTERGER4 = 4
INTERGER6 = 6
NUM_VELOCITY_CALC_SPAN_COUNT: int = 4
SOL = 299792458.0
TIME_DAY_HOUR = 24
TIME_DAY_MINITE = 60
TIME_DAY_SEC = 86400
TIME_MINITE_SEC = 60
__init__(PATH_CEOS_FOLDER: str, POLARIMETORY: str = 'HH', ORBIT_NAME: str = 'A')[source]

Initialize CEOS Format Reader (PALSAR Level 1.1/1.5)

Parameters:

  • PATH_CEOS_FOLDER (str) – CEOS プロダクトが展開されているフォルダ 例: ALPSRP021160650-L1.1/

  • POLARIMETORY (str, optional) – 偏波名 ‘HH’, ‘HV’, ‘VV’, ‘VH’.

  • ORBIT_NAME (str, optional) – 軌道名 ‘A’ or ‘D’.

set_geometory(plot=False, output_json_path: str = None)[source]

観測ジオメトリの設定（L1.0 用実装と同じ変数名で動作）

azayaka.fileformat._write_observation_json(obj, output_path: str)[source]

Write a JSON summary of observation geometry and metadata.

Parameters:

  • obj (object) – Reader instance containing observation attributes.

  • output_path (str) – Output JSON file path.

azayaka.fileformat.check_ceos_polarization_orbit_exists(path_ceos_folder: str, polarimetry: str, orbit_name: str) → None[source]

Check if IMG/LED files matching the given polarization and orbit exist in the CEOS folder. Raises FileNotFoundError if not found.

Parameters:

  • path_ceos_folder – CEOS product folder

  • polarimetry – 偏波名 ‘HH’, ‘HV’, ‘VV’, ‘VH’

  • orbit_name – 軌道名 ‘A’ or ‘D’

Raises:

FileNotFoundError – If the IMG/LED files matching the given polarization and orbit do not exist

azayaka.geocode module

Azayaka:

SAR/InSAR Geocoding Module. This module provides functionalities for geocoding SAR/InSAR outputs data.

Copyright (c) 2026 Syusuke Yasui, Yutaka Yamamoto, and contributors. Licensed under the APGL-3.0 License.

Equation - Condition: No 3.xxx File.

class azayaka.geocode.GRS80[source]

Bases: object

GRS80 (Geodetic Reference System 1980) 測地基準系のパラメータ

AE = 6378137.0
AM = 6370994.431619486
AMSQ = 40589570047726.49
AP = 6356752.314140356
FFACT = -0.006694380022900792
FLAT = 0.003352810681182319
GM = 398600.436
OMFSQ = 0.9933056199770992
class azayaka.geocode.Geocode(sar, dem_path: str = None, dem_bounds: Tuple[float, float, float, float] = None, dem_shape: Tuple[int, int] = None, dem_transform: Affine = None, dem_crs: str | CRS = 'EPSG:4326', buffer_sample: int = 0, look_direction: str = 'R')[source]

Bases: object

SAR/InSAR geocoding workflow.

This class manages DEM loading/creation and provides utilities to map radar-domain data onto geographic coordinates using range-Doppler geometry.

__init__(sar, dem_path: str = None, dem_bounds: Tuple[float, float, float, float] = None, dem_shape: Tuple[int, int] = None, dem_transform: Affine = None, dem_crs: str | CRS = 'EPSG:4326', buffer_sample: int = 0, look_direction: str = 'R')[source]

Initialize the geocoding helper with SAR metadata and DEM settings.

Parameters:

  • sar (object) – SAR reader instance with geometry attributes (orbit, slant range, etc.).

  • dem_path (str, optional) – Path to an existing DEM GeoTIFF. If not provided, an empty DEM is created.

  • dem_bounds (tuple of float, optional) – DEM bounds as (min_lon, min_lat, max_lon, max_lat) in degrees.

  • dem_shape (tuple of int, optional) – DEM shape as (height, width) in pixels.

  • dem_transform (rasterio.Affine, optional) – Raster transform for DEM creation.

  • dem_crs (str or rasterio.crs.CRS, optional) – Coordinate reference system for the DEM.

  • buffer_sample (int, optional) – Margin in radar samples to include when cropping for registration.

  • look_direction (str, optional) – Look direction (“R” or “L”) used for range-Doppler back-geocoding.

static _apply_shift_map(image, h_shift_map, w_shift_map)[source]

Apply per-pixel shifts to an image.

Parameters:

  • image (np.ndarray) – Input image.

  • h_shift_map (np.ndarray) – Per-pixel shift in the vertical direction.

  • w_shift_map (np.ndarray) – Per-pixel shift in the horizontal direction.

Returns:

Shifted image.

Return type:

np.ndarray

static _build_geokey_tags(crs)[source]

Build GeoTIFF key tags from a CRS.

Parameters:

crs (rasterio.crs.CRS or str) – CRS to convert into GeoTIFF tags.

Returns:

GeoTIFF key tags, or empty dict if CRS is invalid.

Return type:

dict

_compute_dem_geometry()[source]

Compute DEM geometry and radar index mappings.

Returns:

  • xyz_dem (np.ndarray) – DEM points in ECEF coordinates with shape (H, W, 3).

  • idx_azimuth (np.ndarray) – Azimuth indices for each DEM pixel.

  • idx_range (np.ndarray) – Range indices for each DEM pixel.

  • idx_invalid (np.ndarray) – Boolean mask of invalid indices.

_compute_scene_corners(max_iter: int = 1000, look_direction: str = 'R')[source]

Compute scene footprint corners via range-Doppler back-geocoding.

Parameters:

  • max_iter (int, optional) – Maximum iterations for the solver.

  • look_direction (str, optional) – Look direction (“R” or “L”).

Returns:

Corner coordinates as (lat, lon) in degrees.

Return type:

list[tuple[float, float]]

classmethod _coregistration_fine_correlation_vectorized(clx_m, clx_s, window_size=32, shift_range_min=-5, shift_range_max=5, stride=8)[source]

Estimate fine coregistration shifts by maximizing correlation.

Parameters:

  • clx_m (np.ndarray) – Main complex image.

  • clx_s (np.ndarray) – Secondary complex image to be shifted.

  • window_size (int, optional) – Window size for correlation computation.

  • shift_range_min (int, optional) – Minimum shift to test (pixels).

  • shift_range_max (int, optional) – Maximum shift to test (pixels).

  • stride (int, optional) – Sampling stride for sparse shift estimation.

Returns:

  • clx_s_reg (np.ndarray) – Coregistered secondary image.

  • coh_best (np.ndarray) – Best correlation map.

  • shift_map (tuple[np.ndarray, np.ndarray]) – (azimuth_shift_map, range_shift_map).

static _correlation_vectorized(clx_m, clx_s, window_size, mean_m_squared=None)[source]

Compute normalized correlation between two complex arrays.

Parameters:

  • clx_m (np.ndarray) – Main complex image.

  • clx_s (np.ndarray) – Secondary complex image.

  • window_size (int) – Window size for local averaging.

  • mean_m_squared (np.ndarray, optional) – Precomputed local mean of |clx_m|^2 for reuse.

Returns:

Correlation (coherence-like) map.

Return type:

np.ndarray

static _fill_nan_values_simple(data, fill_value=0.0)[source]

Replace NaN values with a constant fill value.

Parameters:

  • data (np.ndarray) – Input array possibly containing NaNs.

  • fill_value (float, optional) – Value to replace NaNs with.

Returns:

Array with NaNs replaced.

Return type:

np.ndarray

classmethod _geocode_dem_to_radar_smooth(dem, idx_azimuth, idx_range, num_aperture_sample, num_pixel)[source]

Map DEM into radar coordinates and fill gaps smoothly.

Parameters:

  • dem (np.ndarray) – DEM in geographic grid.

  • idx_azimuth (np.ndarray) – Azimuth index map.

  • idx_range (np.ndarray) – Range index map.

  • num_aperture_sample (int) – Number of azimuth samples.

  • num_pixel (int) – Number of range pixels.

Returns:

  • dem_radar_smooth (np.ndarray) – DEM resampled to radar coordinates.

  • valid_mask (np.ndarray) – Boolean mask of valid samples.

_geocode_radar_to_geo(radar_image, top_az, left_rg, po_shift_azimuth, po_shift_range, use_count=True)[source]

Map a radar image to geographic grid using precomputed indices.

Parameters:

  • radar_image (np.ndarray) – Radar-domain image to geocode.

  • top_az (int) – Top azimuth index for cropping.

  • left_rg (int) – Left range index for cropping.

  • po_shift_azimuth (float) – Azimuth shift to apply before mapping.

  • po_shift_range (float) – Range shift to apply before mapping.

  • use_count (bool, optional) – Whether to average overlapping samples.

Returns:

Geocoded image in DEM grid coordinates.

Return type:

np.ndarray

classmethod _interpolate_with_spline_fixed(dem_sparse, valid_mask, height, width)[source]

Interpolate DEM data using a coarse spline, with fallback interpolation.

Parameters:

  • dem_sparse (np.ndarray) – Sparse DEM in radar coordinates.

  • valid_mask (np.ndarray) – Boolean mask of valid samples.

  • height (int) – Output height.

  • width (int) – Output width.

Returns:

Interpolated DEM array.

Return type:

np.ndarray

_load_or_create_dem()[source]

Load a DEM from disk or create an empty DEM grid.

Returns:

  • dem (np.ndarray) – DEM height array in meters.

  • transform (rasterio.Affine) – Affine transform for the DEM grid.

  • crs (rasterio.crs.CRS or str) – Coordinate reference system for the DEM.

  • bounds (rasterio.coords.BoundingBox) – Bounding box of the DEM.

Raises:

ValueError – If required DEM metadata is missing when dem_path is not provided.

_range_doppler_back_geocode(sat_pos: ndarray, sat_vel: ndarray, slant_range: float, max_iter: int = 50, tol: float = 1e-06, look_direction: str = 'R') → Tuple[float, float][source]

Solve range-Doppler equations to obtain ground coordinates.

Parameters:

  • sat_pos (np.ndarray) – Satellite position (ECEF) as a length-3 vector.

  • sat_vel (np.ndarray) – Satellite velocity (ECEF) as a length-3 vector.

  • slant_range (float) – Slant range distance in meters.

  • max_iter (int, optional) – Maximum number of Newton iterations.

  • tol (float, optional) – Convergence tolerance for the Newton solver.

  • look_direction (str, optional) – Look direction (“R” or “L”) used to select the cross-track direction.

Returns:

  • lat (float) – Geodetic latitude in degrees.

  • lon (float) – Geodetic longitude in degrees.

static _save_geotiff(path, data, crs, transform)[source]

Save a single-band GeoTIFF with basic geokeys.

Parameters:

  • path (str) – Output file path.

  • data (np.ndarray) – 2D array to write.

  • crs (rasterio.crs.CRS or str) – Coordinate reference system.

  • transform (rasterio.Affine) – Affine transform for the raster.

static _save_jpg(path, data)[source]

Save a normalized grayscale JPEG representation of data.

Parameters:

  • path (str) – Output JPEG path.

  • data (np.ndarray) – Input data array.

static _save_multiband_geotiff(path, layers, crs, transform, metadata)[source]

Save a multi-band GeoTIFF with metadata.

Parameters:

  • path (str) – Output file path.

  • layers (list of dict) – Layer descriptors with keys: data, layer_name, process_name, format, scene_id.

  • crs (rasterio.crs.CRS or str) – Coordinate reference system.

  • transform (rasterio.Affine) – Affine transform for the raster.

  • metadata (dict) – Global metadata to store as tags.

Raises:

ValueError – If no layers are provided.

static _simple_interpolation(dem_sparse, valid_mask, height, width)[source]

Fill sparse DEM data using nearest-neighbor interpolation.

Parameters:

  • dem_sparse (np.ndarray) – Sparse DEM in radar coordinates.

  • valid_mask (np.ndarray) – Boolean mask of valid samples.

  • height (int) – Output height.

  • width (int) – Output width.

Returns:

Interpolated DEM array.

Return type:

np.ndarray

geocode(signal: ndarray, phase: ndarray = None, output_intensity_path: str = None, output_phase_path: str = None, register: bool = True)[source]

Geocode radar-domain data to the DEM grid.

Parameters:

  • signal (np.ndarray) – Complex SAR image in radar coordinates.

  • phase (np.ndarray, optional) – Optional phase or complex data to geocode alongside intensity.

  • output_intensity_path (str, optional) – Output GeoTIFF path for intensity products.

  • output_phase_path (str, optional) – Output GeoTIFF path for phase products.

  • register (bool, optional) – Whether to perform DEM-to-image registration before geocoding.

Returns:

Dictionary of geocoded products and registration outputs.

Return type:

dict

Raises:

ValueError – If the input signal is smaller than the SAR geometry.

save_scene_kml(output_kml_path: str, max_iter: int = 1000, look_direction: str = None, include_overlay: bool = True, overlay_size: int = 1024)[source]

Save a scene footprint KML for the current SAR acquisition.

Parameters:

  • output_kml_path (str) – Output path for the KML file.

  • max_iter (int, optional) – Maximum iterations for range-Doppler solver.

  • look_direction (str, optional) – Look direction (“R” or “L”). If None, uses the class default.

  • include_overlay (bool, optional) – Whether to include an intensity overlay.

  • overlay_size (int, optional) – Size of the overlay image (pixels).

Returns:

Scene corners as (lat, lon) in degrees.

Return type:

list[tuple[float, float]]

azayaka.geocode.geocen(lat: float | ndarray, height: float | ndarray) → Tuple[ndarray, ndarray][source]

測地緯度から地心緯度と地心距離を計算する。

測地緯度（地表面の法線が赤道面となす角）を地心緯度（地心から見た角度）に変換し、 地心からの距離を計算します。

Parameters:

  • lat (float or array_like) – 測地緯度 (ラジアン)。スカラー値またはn次元配列。

  • height (float or array_like) – GRS80楕円体からの高度 (m)。latと同じ形状である必要があります。

Returns:

  • latc (ndarray) – 地心緯度 (ラジアン)

  • r (ndarray) – 地心からの距離 (m)

Notes

この関数は、楕円体上の点から地心座標系への変換において、 緯度の補正と実際の地心距離を計算します。

azayaka.geocode.geoxyz(lat: float | ndarray, lon: float | ndarray, height: float | ndarray) → Tuple[ndarray, ndarray][source]

測地座標から地心直交座標（ECEF）への変換。

GRS80測地基準系における測地緯度、経度、楕円体高から、 地球中心を原点とする地心直交座標系（Earth Centered Earth Fixed: ECEF）の XYZ座標を計算します。

Parameters:

  • lat (float or array_like) – 測地緯度 (ラジアン)。-π/2 ≤ lat ≤ π/2

  • lon (float or array_like) – 測地経度 (ラジアン)。-π ≤ lon ≤ π

  • height (float or array_like) – GRS80楕円体からの高度 (m)

Returns:

  • xyz (ndarray) – 地心直交座標 [X, Y, Z] (m)。 - X軸: グリニッジ子午線と赤道の交点方向 - Y軸: 東経90度と赤道の交点方向 - Z軸: 地球の自転軸（北極方向が正）

  • r (ndarray) – 地心からの距離 (m)

See also

xyz2geo

逆変換（地心直交座標から測地座標への変換）

azayaka.geocode.polcar(lat: float | ndarray, lon: float | ndarray, r: float | ndarray) → ndarray[source]

極座標（球座標）から直交座標への変換。

地心距離と地心緯度・経度から、地心直交座標系（ECEF: Earth Centered Earth Fixed） のXYZ座標を計算します。

Parameters:

  • lat (float or array_like) – 地心緯度 (ラジアン)

  • lon (float or array_like) – 経度 (ラジアン)

  • r (float or array_like) – 地心からの距離 (m)

Returns:

xyz – 地心直交座標 [X, Y, Z] (m)。 入力が配列の場合、最後の次元が3要素の座標になります。 例：入力が(M, N)の場合、出力は(M, N, 3)

Return type:

ndarray

azayaka.geocode.xyz2geo(xyz: ndarray | Tuple[float, float, float]) → Tuple[ndarray, ndarray, ndarray][source]

地心直交座標（ECEF）から測地座標への逆変換。

Parameters:

xyz (array_like) – 地心直交座標 [X, Y, Z] (m)。 形状は(…, 3)である必要があります。

Returns:

  • lat (ndarray) – 測地緯度 (ラジアン)

  • lon (ndarray) – 測地経度 (ラジアン)

  • height (ndarray) – 楕円体高 (m)

Notes

Bowringの反復法を使用した逆変換アルゴリズムを実装しています。

azayaka.interferometry module

Azayaka:

Interferometry SAR Module. This module provides functionalities for SAR interferometric processing and filtering.

Copyright (c) 2026 Syusuke Yasui, Yutaka Yamamoto, and contributors. Licensed under the APGL-3.0 License.

Equation - Condition: No 1.xxx File.

class azayaka.interferometry.Interferometry(main, sub)[source]

Bases: object

SAR interferometry processing pipeline.

Provides coregistration, interferogram generation, filtering, and geocoding utilities for a main/sub SLC pair.

__init__(main, sub)[source]

Initialize the interferometry workflow with main and sub SLC data.

Parameters:

  • main (object) – Primary SAR reader instance.

  • sub (object) – Secondary SAR reader instance.

_align_signals_to_main() → None[source]

Align main and sub signals by cropping to the common shape.

Return type:

None

static _apply_geocode_registration(geocoder: Geocode, radar_image: ndarray, registration: Dict) → ndarray[source]

Apply registration parameters and geocode a radar image.

Parameters:

  • geocoder (Geocode) – Geocode helper instance.

  • radar_image (np.ndarray) – Radar-domain image to transform.

  • registration (dict) – Registration metadata from _prepare_geocode_registration.

Returns:

Geocoded image.

Return type:

np.ndarray

static _apply_geocode_registration_cropped(geocoder: Geocode, radar_crop: ndarray, registration: Dict) → ndarray[source]

Apply registration to a pre-cropped radar image and geocode.

Parameters:

  • geocoder (Geocode) – Geocode helper instance.

  • radar_crop (np.ndarray) – Cropped radar-domain image.

  • registration (dict) – Registration metadata from _prepare_geocode_registration.

Returns:

Geocoded image.

Return type:

np.ndarray

static _coherence_vectorized(clx_m: ndarray, clx_s: ndarray, window_size: int, mean_m_squared: ndarray | None = None) → ndarray[source]

Compute local interferometric coherence.

Parameters:

  • clx_m (np.ndarray) – Main complex image.

  • clx_s (np.ndarray) – Secondary complex image.

  • window_size (int) – Window size for local averaging.

  • mean_m_squared (np.ndarray, optional) – Precomputed mean of |clx_m|^2 for reuse.

Returns:

Coherence map.

Return type:

np.ndarray

_compute_topography_phase(dem_radar: ndarray) → ndarray[source]

Simulate topography phase for the full radar grid.

Parameters:

dem_radar (np.ndarray) – DEM resampled to radar coordinates.

Returns:

Complex topography phase screen.

Return type:

np.ndarray

_compute_topography_phase_cropped(dem_radar_crop: ndarray, top_az: int, left_rg: int) → ndarray[source]

Simulate topography phase for a cropped radar window.

Parameters:

  • dem_radar_crop (np.ndarray) – Cropped DEM in radar coordinates.

  • top_az (int) – Top azimuth index of the crop in the full grid.

  • left_rg (int) – Left range index of the crop in the full grid.

Returns:

Complex topography phase screen for the crop.

Return type:

np.ndarray

static _convolve2d(image: ndarray, kernel: ndarray) → ndarray[source]

Perform a valid 2D convolution using stride tricks.

Parameters:

  • image (np.ndarray) – Input image.

  • kernel (np.ndarray) – Convolution kernel.

Returns:

Convolved output.

Return type:

np.ndarray

classmethod _convolve2d_safe(image: ndarray, kernel: ndarray, boundary: str = 'edge') → ndarray[source]

Convolve with padding to preserve size.

Parameters:

  • image (np.ndarray) – Input image.

  • kernel (np.ndarray) – Convolution kernel.

  • boundary (str, optional) – Padding mode. If empty, no padding is applied.

Returns:

Convolved output.

Return type:

np.ndarray

_coregister_slc(slc_main: ndarray, slc_sub: ndarray, fine: bool = True, coherence_window: int = 4, fine_shift_range: int = 2, fine_stride: int = 1, coarse_downsample: int = 1) → Tuple[ndarray, ndarray, Tuple[float, float], Tuple[ndarray, ndarray]][source]

Coregister a secondary SLC to the main SLC.

Parameters:

  • slc_main (np.ndarray) – Main SLC image.

  • slc_sub (np.ndarray) – Secondary SLC image.

  • fine (bool, optional) – Whether to perform fine coregistration.

  • coherence_window (int, optional) – Window size for coherence computation.

  • fine_shift_range (int, optional) – Search range (pixels) for fine shifts.

  • fine_stride (int, optional) – Stride for fine shift search.

  • coarse_downsample (int, optional) – Downsample factor for coarse phase correlation.

Returns:

  • slc_sub_reg (np.ndarray) – Coregistered secondary SLC.

  • coherence_reg (np.ndarray) – Coherence after registration.

  • coarse_shift (tuple[float, float]) – Coarse (range, azimuth) shift estimate.

  • shift_map (tuple[np.ndarray, np.ndarray]) – Fine per-pixel shift maps (azimuth, range).

classmethod _coregistration_fine_coherence_vectorized(clx_m: ndarray, clx_s: ndarray, window_size: int = 4, shift_range_min: int = -2, shift_range_max: int = 2, stride: int = 1) → Tuple[ndarray, ndarray, Tuple[ndarray, ndarray]][source]

Estimate fine shifts by maximizing coherence.

Parameters:

  • clx_m (np.ndarray) – Main complex image.

  • clx_s (np.ndarray) – Secondary complex image.

  • window_size (int, optional) – Window size for coherence computation.

  • shift_range_min (int, optional) – Minimum shift (pixels).

  • shift_range_max (int, optional) – Maximum shift (pixels).

  • stride (int, optional) – Sampling stride for sparse search.

Returns:

  • clx_s_reg (np.ndarray) – Coregistered secondary image.

  • coh_best (np.ndarray) – Best coherence map.

  • shift_map (tuple[np.ndarray, np.ndarray]) – (azimuth_shift_map, range_shift_map).

static _create_averaging_kernel(size: Tuple[int, int]) → ndarray[source]

Create a normalized averaging kernel.

Parameters:

size (tuple[int, int]) – Kernel size.

Returns:

Averaging kernel.

Return type:

np.ndarray

static _earth_sign(x_main, y_main, x_sub, y_sub, orbit_name: str, sign: int = 1) → int[source]

Determine the sign for baseline orientation.

Parameters:

  • x_main (np.ndarray) – Main satellite ECEF coordinates.

  • y_main (np.ndarray) – Main satellite ECEF coordinates.

  • x_sub (np.ndarray) – Sub satellite ECEF coordinates.

  • y_sub (np.ndarray) – Sub satellite ECEF coordinates.

  • orbit_name (str) – Orbit direction label (e.g., “A”, “D”).

  • sign (int, optional) – Initial sign value.

Returns:

Sign adjusted by orbit direction and relative longitude.

Return type:

int

classmethod _goldstein_filter_patch(patch: ndarray, alpha: float, filter_kernel: ndarray) → ndarray[source]

Apply Goldstein phase filtering to a single patch.

Parameters:

  • patch (np.ndarray) – Complex input patch.

  • alpha (float) – Goldstein filter exponent.

  • filter_kernel (np.ndarray) – Smoothing kernel in frequency domain.

Returns:

Filtered patch.

Return type:

np.ndarray

classmethod _goldstein_phase_filter(image: ndarray, alpha: float = 0.4, patch_size: int = 64, step: int = 8, filter_size: int = 3) → ndarray[source]

Apply Goldstein phase filtering over an image.

Parameters:

  • image (np.ndarray) – Complex interferogram or phase image.

  • alpha (float, optional) – Goldstein exponent.

  • patch_size (int, optional) – Patch size for local filtering.

  • step (int, optional) – Step size for patch traversal.

  • filter_size (int, optional) – Size of averaging kernel for smoothing.

Returns:

Filtered image.

Return type:

np.ndarray

_init_baseline_geometry(baseline_offset=0.0) → None[source]

Compute baseline geometry between main and sub orbits.

Parameters:

baseline_offset (float, optional) – Offset added to the perpendicular baseline component.

Return type:

None

static _multilook_filter(image: ndarray, looks_azimuth: int, looks_range: int) → ndarray[source]

Apply multilook averaging in azimuth and range.

Parameters:

  • image (np.ndarray) – Input image.

  • looks_azimuth (int) – Number of looks in azimuth.

  • looks_range (int) – Number of looks in range.

Returns:

Filtered image.

Return type:

np.ndarray

static _pad_singlechannel_image(image: ndarray, kernel_shape: Tuple[int, int], boundary: str) → ndarray[source]

Pad a single-channel image for convolution.

Parameters:

  • image (np.ndarray) – Input image.

  • kernel_shape (tuple[int, int]) – Kernel shape.

  • boundary (str) – Padding mode passed to np.pad.

Returns:

Padded image.

Return type:

np.ndarray

static _pad_to_shape(data: ndarray, target_shape: Tuple[int, int], fill_value) → ndarray[source]

Pad an array to a target shape.

Parameters:

  • data (np.ndarray) – Input array.

  • target_shape (tuple[int, int]) – Desired output shape.

  • fill_value (scalar) – Fill value for padded regions.

Returns:

Padded array.

Return type:

np.ndarray

_prepare_geocode_registration(geocoder: Geocode, signal: ndarray, dem_coreg_window_size: int, dem_coreg_shift_range: int, dem_coreg_stride: int)[source]

Compute DEM-based registration parameters for geocoding.

Parameters:

  • geocoder (Geocode) – Geocode helper instance.

  • signal (np.ndarray) – Radar-domain signal used for registration.

  • dem_coreg_window_size (int) – Window size for DEM-to-image correlation.

  • dem_coreg_shift_range (int) – Search range (pixels) for DEM registration.

  • dem_coreg_stride (int) – Sampling stride for the correlation search.

Returns:

Registration metadata including crop window and shift maps.

Return type:

dict

static _save_histogram_jpg(values: ndarray, path: str, bins: int = 256, vmin: float = 0.0, vmax: float = 1.0, threshold: float | None = None) → None[source]

Save a histogram plot to JPEG.

Parameters:

  • values (np.ndarray) – Input values for histogram.

  • path (str) – Output file path.

  • bins (int, optional) – Number of histogram bins.

  • vmin (float, optional) – Minimum histogram range.

  • vmax (float, optional) – Maximum histogram range.

  • threshold (float, optional) – Optional threshold indicator to draw on the histogram.

static _save_jpg(image: ndarray, path: str, vmin: float | None = None, vmax: float | None = None, cmap: str | None = None) → None[source]

Save an image to JPEG with optional colormap.

Parameters:

  • image (np.ndarray) – Input image.

  • path (str) – Output file path.

  • vmin (float, optional) – Minimum value for normalization.

  • vmax (float, optional) – Maximum value for normalization.

  • cmap (str, optional) – Matplotlib colormap name. If None, saves grayscale.

process(output_dir: str, dem_path: str | None = None, dem_bounds: Tuple[float, float, float, float] | None = None, dem_shape: Tuple[int, int] | None = None, dem_transform=None, dem_crs: str = 'EPSG:4326', buffer_sample: int = 0, look_direction: str = 'R', output_prefix: str = 'interferometry', fine_registration: bool = True, coherence_window: int = 4, fine_shift_range: int = 2, fine_stride: int = 1, multilook_azimuth: int = 3, multilook_range: int = 4, goldstein_alpha: float = 0.4, goldstein_patch_size: int = 64, goldstein_step: int = 16, goldstein_filter_size: int = 3, coherence_threshold_quantile: float | None = 0.3333333333333333, coherence_histogram_threshold: float | None = None, dem_coreg_window_size: int = 128, dem_coreg_shift_range: int = 1, dem_coreg_stride: int = 2, slc_coreg_coarse_downsample: int = 1, sub_buffer: int = 1000) → Dict[str, str][source]

Run the full interferometry processing pipeline and save outputs.

Parameters:

  • output_dir (str) – Output directory for GeoTIFF/JPEG products.

  • dem_path (str, optional) – Path to DEM GeoTIFF.

  • dem_bounds (tuple[float, float, float, float], optional) – DEM bounds (min_lon, min_lat, max_lon, max_lat).

  • dem_shape (tuple[int, int], optional) – DEM shape (height, width).

  • dem_transform (rasterio.Affine, optional) – DEM affine transform.

  • dem_crs (str, optional) – DEM CRS (e.g., “EPSG:4326”).

  • buffer_sample (int, optional) – Cropping buffer in radar samples.

  • look_direction (str, optional) – Look direction (“R” or “L”).

  • output_prefix (str, optional) – Prefix for output filenames.

  • fine_registration (bool, optional) – Enable fine SLC coregistration.

  • coherence_window (int, optional) – Window size for coherence estimation.

  • fine_shift_range (int, optional) – Search range for fine SLC shifts.

  • fine_stride (int, optional) – Stride for fine shift search.

  • multilook_azimuth (int, optional) – Multilook factor in azimuth.

  • multilook_range (int, optional) – Multilook factor in range.

  • goldstein_alpha (float, optional) – Goldstein filter exponent.

  • goldstein_patch_size (int, optional) – Patch size for Goldstein filtering.

  • goldstein_step (int, optional) – Patch step for Goldstein filtering.

  • goldstein_filter_size (int, optional) – Filter kernel size for Goldstein filtering.

  • coherence_threshold_quantile (float, optional) – Quantile to threshold coherence.

  • coherence_histogram_threshold (float, optional) – Explicit threshold for coherence masking.

  • dem_coreg_window_size (int, optional) – Window size for DEM coregistration.

  • dem_coreg_shift_range (int, optional) – Shift range for DEM coregistration.

  • dem_coreg_stride (int, optional) – Stride for DEM coregistration.

  • slc_coreg_coarse_downsample (int, optional) – Downsample factor for coarse SLC coregistration.

  • sub_buffer (int, optional) – Buffer margin for sub-image cropping.

Returns:

Mapping of output product names to file paths.

Return type:

dict[str, str]

azayaka.processing module

Azayaka:

SAR/InSAR Processing Handler Module. Summary module for various SAR/InSAR data processing tasks.

Copyright (c) 2026 Syusuke Yasui, Yutaka Yamamoto, and contributors. Licensed under the APGL-3.0 License.

azayaka.utils module

Azayaka utility functions.

azayaka.utils.save_scene_kml(geocoder, output_kml_path: str, max_iter: int = 1000, look_direction: str = None, include_overlay: bool = True, overlay_size: int = 1024, xyz2geo_func: Callable[[ndarray], Tuple[ndarray, ndarray, ndarray]] = None)[source]

Export the SAR scene footprint to a KML file.

Parameters:

  • geocoder (Geocode) – Geocode instance containing SAR geometry and orbit information.

  • output_kml_path (str) – Output path for the KML file.

  • max_iter (int, optional) – Maximum iterations for the range-Doppler back-geocoding solver.

  • look_direction (str, optional) – Look direction override (“R” or “L”). If None, uses the geocoder setting.

  • include_overlay (bool, optional) – Whether to embed a downsampled intensity overlay.

  • overlay_size (int, optional) – Output overlay size in pixels for both width and height.

  • xyz2geo_func (callable, optional) – Function to convert ECEF XYZ to (lat, lon, height) in radians/meters.

Returns:

Scene corners as (lat, lon) in degrees, ordered around the footprint.

Return type:

list[tuple[float, float]]

Raises:

ValueError – If xyz2geo_func is not provided or computed corners are out of bounds.

Module contents

Azayaka:

SAR/InSAR Python Package. It includes various utilities for SAR/InSAR data processing and analysis.

QGIS Plugin “Azayaka” is also provided for easy access to some functionalities.

Copyright (c) 2026 Syusuke Yasui, Yutaka Yamamoto, and contributors. Licensed under the APGL-3.0 License.