Next-generation radio interferometers such as the Square Kilometre Array Observatory (SKAO) will produce data at petabyte scales. A central challenge is to image extremely faint, diffuse radio emission (e.g. the cosmic web, galaxy-cluster halos, or subtle solar features) amid much brighter compact sources. Traditional radio astronomy imaging (e.g. CLEAN) treats the sky as discrete sources and often requires lengthy deconvolution, which can leave residual artefacts around extended emission. In contrast, the Bluebild algorithm offers a fresh approach by treating interferometric imaging as a continuous inverse problem. Bluebild applies a functional principal-component decomposition to the measured visibilities, effectively splitting the sky into a sequence of orthogonal â energy levelsâ that are recombined into a least-squares sky map. In practice, this means bright sources are captured by the leading modes, while fainter diffuse structures appear in lower modes.

In this thesis, a high-performance implementation of Bluebildâ called Bluebild Imaging Plus Plus (BIPP)â is introduced and validated. BIPP is written in C++ with GPU acceleration and interfaces to Python, and it exploits 3-D non-uniform fast Fourier transforms (NUFFTs) to image the sky directly on the celestial sphere. The software was tested on both simulated SKA datasets and real telescope data (e.g. from LOFAR and the MWA). These tests show that BIPP produces accurate wide-field images with computational cost comparable to standard pipelines (CASA/WSClean). Importantly, BIPPâ s energy-level decomposition often exposes faint diffuse emission before any deconvolution steps. For example, after running BIPP, one can â peelâ bright sources by removing the top principal modes, yielding residual data that cleanly reveal weak, extended structures. In effect, BIPP enables the separation of â bright foregroundâ emission from the underlying diffuse background.

One striking result obtained with BIPP is the serendipitous detection of solar limb brightening in meter-wavelength data from the MWA. The thesis shows that quiet-Sun observations imaged with BIPP display the solar disk edges as unusually bright, confirming a long-sought feature of solar radio emission. A companion simulation (with a known limb-brightening component) was processed by BIPP, which successfully split the Sunâ s central disk emission from the diffuse limb emission. This demonstrates that Bluebild/BIPP can reveal subtle astrophysical phenomena in practice. More generally, the method is well suited to science cases like quiet-Sun imaging, Epoch-of-Reionisation studies, or detection of the cosmic web, where faint diffuse signals must be separated from strong foregrounds.

Looking ahead, the thesis argues that Bluebild/BIPP could play a key role in SKA-era data processing. By embedding the eigen-decomposition approach into the calibration and imaging pipeline, bright contaminating sources (e.g. radio galaxies, radio frequency interference, or solar emission) could be filtered out in visibility space, leaving a cleaner dataset for imaging faint features. In summary, this research provides a practical new toolkit for interferometric imaging: BIPP demonstrates that modern HPC and statistical methods can recover extended sky emission with high fidelity. As next-generation arrays come online, these tools promise to expand the scientific reach of radio astronomy by unveiling diffuse signals that are otherwise hidden.