Spaceborne LiDAR systems are crucial for Earth observation but face hardware constraints, thus limiting resolution and data processing. We propose integrating compressed sensing and diffusion generative models to reconstruct high-resolution satellite LiDAR data within the Hyperheight Data Cube (HHDC) framework. Using a randomized illumination pattern in the imaging model, we achieve efficient sampling and compression, reducing the onboard computational load and optimizing data transmission. Diffusion models then reconstruct detailed HHDCs from sparse samples on Earth. To ensure reliability despite lossy compression, we analyze distortion metrics for derived products like Digital Terrain and Canopy Height Models and evaluate the 3D reconstruction accuracy in waveform space. We identify image quality assessment metrics—ADD_GSIM, DSS, HaarPSI, PSIM, SSIM4, CVSSI, MCSD, and MDSI—that strongly correlate with subjective quality in reconstructed forest landscapes. This work advances high-resolution Earth observation by combining efficient data handling with insights into LiDAR imaging fidelity.

One of the continuous challenges related to the growing popularity of mobile devices and embedded systems with limited memory and computational power is the development of relatively fast methods for real-time image and video analysis. One such example is Optical Character Recognition (OCR), which is usually too complex for such devices. Considering that images captured by cameras integrated into mobile devices may be acquired in uncontrolled lighting conditions, some quality issues related to non-uniform illumination may affect the image binarization results and further text recognition results. The solution proposed in this paper is related to a significant reduction in the computational burden, preventing the necessity of full text recognition. Conducting only the initial image binarization using various thresholding methods, the computation of the mutual similarities of binarization results is proposed, making it possible to build a simple model of binary image quality for a fast prediction of the OCR results’ quality. The experimental results provided in the paper obtained for the dataset of 1760 images, as well as the additional verification for a larger dataset, confirm the high correlation of the proposed quality model with text recognition results.

Spaceborne LiDAR reconstructions from compressed HyperHeight Data Cubes (HHDCs) are increasingly obtained with diffusion models, but reliable quality assessment (QA) tools tailored to LiDAR-derived rasters are lacking. This paper closes that gap by (i) building a subjective database for HHDC percentile products—Canopy Height Models (CHMs), median height maps, and Digital Terrain Models (DTMs)—with 342 reconstruction/reference pairs and Mean Opinion Scores (MOS) collected from multiple observers in both pseudo-colored and grayscale renderings; (ii) benchmarking 70+ full-reference image quality assessment (FR-IQA) metrics against MOS; and (iii) proposing a compact nonlinear fusion of top performers. Data-driven FR IQA methods (DISTS, TOPIQ trained on PIPAL, LPIPS/LPIPS+/ST-LPIPS) exhibit the best correlation with human opinion on LiDAR-derived rasters, achieving absolute PLCC/SROCC values of around 0.85–0.86 on aggregate. A simple weighted sum–product fusion of five complementary metrics further increases alignment to PLCC/SROCC of 0.9221/0.9204 across all percentiles and modalities. Cross-modal consistency of the subjective scores is high (e.g., grayscale vs. pseudo-color for the same percentile), while pseudo-colored DTMs exhibit the most outliers, indicating colormap- and homogeneity-induced instability in some metrics. These findings enable practical QA for generative LiDAR reconstruction: the fused metric serves as a reliable MOS surrogate for evaluation and as a perceptual loss to regularize diffusion-based reconstructions. The released dataset and benchmarks provide a reference for future LiDAR-specific IQA and extensions to tensor-level HHDC assessment.

In this paper, experimental results concerning the application of no-reference image quality metrics for automatic quality assessment of reconstructed satellite light detection and ranging (LiDAR) data are discussed. Since satellite LiDARs provide much sparser data compared to airborne systems and their sending to Earth requires additional compression, the received data should be reconstructed. For this purpose, generative neural networks can be applied, including generative diffusion models; however, they require the choice of appropriate parameters, and the reconstruction results should be reliably assessed. Although the physical representation of data is related to the number of received photons, considering the novel data representation based on hyper-height data cubes, it is possible to analyze individual percentiles, representing canopy height models or digital terrain models, as grayscale images. Therefore, their quality may be evaluated using image quality assessment metrics. Due to the lack of reference data, in the application considered, only no-reference metrics may be efficiently applied, even though they have lower correlation with subjective quality perception compared to full-reference metrics. Therefore, the idea of designing combined no-reference metrics optimized for the highest correlation with subjective quality scores is discussed, along with the presentation of the obtained correlation results.