add snr threshold filtering

src/depiction/spectrum/peak_filtering/filter_by_snr_threshold.py

@@ -0,0 +1,33 @@
+from dataclasses import dataclass
+
+import numpy as np
+import scipy
+import scipy.signal
+from numpy.typing import NDArray
+
+from depiction.spectrum.peak_filtering import PeakFilteringType
+
+
+@dataclass
+class FilterBySnrThreshold(PeakFilteringType):
+    """Implements SNR threshold based on a median absolute deviation (MAD) estimate of the noise level."""
+    snr_threshold: float
+    window_size: int = 10
+
+    def filter_peaks(
+        self,
+        spectrum_mz_arr: NDArray[float],
+        spectrum_int_arr: NDArray[float],
+        peak_mz_arr: NDArray[float],
+        peak_int_arr: NDArray[float],
+    ) -> tuple[NDArray[float], NDArray[float]]:
+        noise_level = self._estimate_noise_level(signal=spectrum_int_arr)
+        snr = spectrum_int_arr / noise_level
+        selection = snr > self.snr_threshold
+        return peak_mz_arr[selection], peak_int_arr[selection]
+
+    def _estimate_noise_level(self, signal: NDArray[float]) -> NDArray[float]:
+        """Estimates the noise level in the signal using median absolute deviation (MAD)."""
+        # TODO window size again should be adjustable in different units -> centralize this common functionality
+        filtered_signal = scipy.signal.medfilt(signal, kernel_size=self.window_size)
+        return np.abs(signal - filtered_signal)