Bishr Omer

Xi'an, China

I am an MSc researcher at Northwestern Polytechnical University (NPU), Xi’an, China, supervised by Associate Professor Xu Li. My research focuses Vision-Language Models (VLMs), Multimodal Learning, and Computer Vision. I am particularly interested in building intelligent systems that integrate visual and semantic information for image understanding, quality assessment, and scientific imaging applications.

I received my BSc in Electrical and Communication Engineering from Blue Nile University, Sudan, where I built a foundation in signal processing, digital communications, and embedded systems.

Research Interests: Vision-Language Models · Multimodal Learning · Computer Vision · Blind Image Quality Assessment · 3D Gaussian Splatting Forensics · Scientific Image Analysis.

My recent work bridges traditional image quality assessment and modern foundation models by developing a distortion-aware multimodal framework that combines handcrafted statistical features with frozen Vision-Language Models. This work is submitted to Pattern Recognition (Elsevier, 2026) and available as an arXiv preprint.

My work has been published in Sensors (MDPI, 2026) and recognized with a Best Paper Award at ISCTech 2025.

Research vision

I aim to develop multimodal AI systems that:

Learn robust visual representations through language supervision.
Bridge low-level vision and foundation models.
Understand scientific and remote sensing imagery.
Improve the reliability and interpretability of AI systems.
Advance multimodal reasoning for real-world applications.

News

2026: Submitted “Distortion-Aware Fusion of Statistical and Vision-Language Features for Blind Image Quality Assessment” to Pattern Recognition.
2026: Published “A No-Reference Multivariate Gaussian-Based Spectral Distortion Index for Pansharpened Images” in Sensors.
2025: Received the Best Paper Award at ISCTech 2025.

selected publications

SPIC
Parameter-Efficient Adaptation of a Multi-Stream Vision-Language Framework for Blind Image Quality Assessment

Bishr O. Adam and Xu Li

Signal Processing: Image Communication, 2026

Under review

Abs arXiv Bib Code

Blind image quality assessment (BIQA) increasingly relies on frozen vision-language model (VLM) embeddings, which are efficient but leave open how much accuracy is sacrificed without backbone adaptation and when such adaptation helps. We develop an efficient BIQA framework that fuses a 138-dimensional NSS descriptor with frozen SigLIP and CLIP-H embeddings through a lightweight regression head, and adapt its SigLIP stream with parameter-efficient Low-Rank Adaptation (LoRA), training only 0.23% of its parameters. Evaluating both frozen and adapted models across six datasets under image-level and reference-level protocols, we show that image-level splitting inflates frozen-feature SROCC by up to 0.44 and masks wide variation in true difficulty that only reference-level evaluation reveals. Adaptation recovers performance in proportion to this difficulty, with the largest gains where frozen features generalize poorly (up to +0.328 SROCC on TID2013).
@article{adam2026peft, title = {Parameter-Efficient Adaptation of a Multi-Stream Vision-Language Framework for Blind Image Quality Assessment}, author = {Adam, Bishr O. and Li, Xu}, journal = {Signal Processing: Image Communication}, publisher = {Elsevier}, year = {2026}, note = {Under review}, }

Sensors

A No-Reference Multivariate Gaussian-Based Spectral Distortion Index for Pansharpened Images

Bishr O.A. Adam, Xu Li, J. Wu, and 1 more author

Sensors, 2026

DOI Bib HTML

@article{adam2026sensors,
  title = {A No-Reference Multivariate Gaussian-Based Spectral Distortion Index for Pansharpened Images},
  author = {Adam, Bishr O.A. and Li, Xu and Wu, J. and Hao, X.},
  journal = {Sensors},
  volume = {26},
  number = {1002},
  year = {2026},
  publisher = {MDPI},
  doi = {10.3390/s26031002},
}

ISCTech

Multivariate Gaussian-Based No-Reference Quality Assessment for Spatial Distortion in Pansharpened Images

Bishr O.A. Adam, Xu Li, Y. Wang, and 1 more author

In Proceedings of the 13th International Conference on Information Systems and Computing Technology (ISCTech 2025), 2027

Best Paper DOI Bib HTML

Best Paper Award

@inproceedings{adam2025isctech,
  title = {Multivariate Gaussian-Based No-Reference Quality Assessment for Spatial Distortion in Pansharpened Images},
  author = {Adam, Bishr O.A. and Li, Xu and Wang, Y. and Yang, X.},
  booktitle = {Proceedings of the 13th International Conference on Information Systems and Computing Technology (ISCTech 2025)},
  series = {Lecture Notes in Networks and Systems},
  volume = {1715},
  publisher = {Springer},
  address = {Singapore},
  year = {2027},
  doi = {10.1007/978-981-95-4448-6_21},
}