Title:Enhanced long-duration gravitational-wave transient sources search pipeline with denoising and tree clustering algorithms

Abstract:We present a two-stage upgrade to the PySTAMPAS pipeline that boosts the search for long-duration (10 to 10^3 s) transients in gravitational-wave detector data. First, a denoising scheme combines complex 2D wavelet shrinkage with an adaptive pixel threshold to suppress noise while retaining signal power. Second, a KDTree nearest-neighbour algorithm clusters surviving pixels in O(log n) time, replacing the standard clustering approach. Tests with one week of LIGO O3b data show a large reduction in false-alarm rate and up to a factor-of-two improvement in search sensitivity. The computational time is also significantly reduced. These gains extend the sensitivity of all-sky, all-time searches to weaker and shorter transients, enabling rapid and deeper analyses in forthcoming LIGO-Virgo-KAGRA observation campaigns.