feat: support whittaker, cubic, and bernstein baseline algorithms

[hamed-musallam]

• whittaker options: lambda, scale, maxIterations
• cubic: no options
• bernstein options: maxIterations, tolerance, factorStd, learningRate, minWeight

close #3969

[cloudflare-workers-and-pages]

Deploying nmrium with    Cloudflare Pages

Latest commit:	2ae875b
Status:	✅  Deploy successful!
Preview URL:	https://76eb700e.nmrium.pages.dev
Branch Preview URL:	https://support-new-baseline-methods.nmrium.pages.dev

View logs

[hamed-musallam]

@lpatiny @jobo322

The new baseline correction filter options do not have enough room in the NMRium header. Which options are important to stay there? , FYI, all options will exist in the processing panel as well.

@jobo322
The 'bernstein' crashes when tested. Could you have a look?

[hamed-musallam]

@jobo322

Are there any constraints on the min and max values for each option?

[targos]

@hamed-musallam

The 'bernstein' crashes when tested. Could you have a look?

degree is a mandatory option in the Bernstein algorithm:

[jobo322]

CubicSpline(data, options)
Robust iterative cubic spline baseline correction that uses anchor points and iterative clipping to avoid pulling the baseline up through peaks.

Options:
zones (array): Array of {from, to} ranges defining baseline regions to prioritize for anchor selection.

header options:
numAnchors (number, default: Math.ceil(signal.length / 500), clamped to [8, 512]): Number of uniformly-spaced anchor points to seed the algorithm.
maxIterations (number, default: 10): Maximum number of clipping iterations.
noiseThreshold (number, default: 1.0): Multiplier for noise level to determine clipping threshold. Smaller values are more aggressive.

Only right panel:
tolerance (number, default: 1e-6): Convergence tolerance as fraction of noise level.
noiseLevel (number, optional): Override automatic noise estimation with a fixed value.
noisePercentile (number, default: 10): Percentile used for automatic noise estimation from signal differences.

[jobo322]

Bernstein(data, degree, options)
Iterative baseline correction using Bernstein polynomials with adaptive weighting.

Options:

zones (array): Array of {from, to} ranges defining baseline regions.

header options:
degree (number, default: 3): max value should be 25 a min value is 3 I guess
factorStd (number, default: 3): Factor to scale residuals for weight calculation.
learningRate (number, default: 0.3): Learning rate for iterative weight updates.
maxIterations (number, default: 100): Maximum number of iterations.

Only right panel:
tolerance (number, default: 1e-6): Convergence tolerance for RMS change.
logger (Logger, optional): Logger instance for debug information.

[jobo322]

• Whittaker(data, options)
  Baseline correction using Whittaker smoothing with iterative reweighting.

Options:

zones (array): Array of {from, to} ranges defining baseline regions.

header options:
learningRate (number, default: 0.2): Weight update factor relative to standard deviation.
lambda (number, default: 2000): Smoothing parameter for the Whittaker smoother.
maxIterations (number, default: 20): Maximum number of iterations.

Only right panel:
tolerance (number, default: 1e-3): Convergence tolerance for baseline change.