NRGAN-HFF: Noise-Resilient GAN with Synergistic Feature Fusion for Reliable and Performable Image Retrieval

doi:10.23940/ijpe.26.05.p4.263273

Abstract

Abstract: CBIR systems work well with clean query images but are highly degraded by real-world noise such as Gaussian blur, salt-and-pepper noise, compression artifacts, and haze. In this paper, NRGAN-HFF (Noise-Resilient GAN with Hybrid Feature Fusion) is proposed as a new CBIR framework that addresses noisy query-image retrieval. The framework incorporates a DnCNN-based Generative Adversarial Network (GAN) to do adaptive query denoising as well as a Synergistic Hybrid Feature Fusion (HFF) module that fuses handcrafted descriptors (Color Moments, LBP, HOG, SIRF) with deep CNN features (fine-tuned VGG16 and pre-trained ResNet50), and then features are selected by LightGBM as well as dimensionality reduction by PCA. A standardized Noisy Query Generation Pipeline (NQGP) effectively uses four types of noise at three intensity levels, which result in 12 variants of corruption for each image to be evaluated objectively. The experiments that were performed on a merged dataset of 19,680 images of Caltech-101, Stanford-40 Actions, and Corel-1K also prove that NRGAN-HFF has an average Noise Robustness Score (NRS) of 94.1%, regardless of the type and severity of noise, which is evidence that the framework is dual-performable with respect to both retrieval and operational resilience in adverse conditions.

Key words: image retrieval, noise robustness, generative adversarial network (GAN), image denoising, system performability, retrieval reliability

Ashish Jain and Sudeep Varshney. NRGAN-HFF: Noise-Resilient GAN with Synergistic Feature Fusion for Reliable and Performable Image Retrieval [J]. Int J Performability Eng, 2026, 22(5): 263-273.

Add to citation manager EndNote|Reference Manager|ProCite|BibTeX|RefWorks

References

[1] Haralick R.M., Shanmugam K., andDinstein I.H., 2007. Textural features for image classification. IEEE Transactions on Systems, Man, and Cybernetics, (6), pp. 610-621.
[2] Srivastava D., Singh S.S., Rajitha B., Verma M., Kaur M., andLee H.N., 2023. Content-based image retrieval: A survey on local and global features selection, extraction, representation, and evaluation parameters.IEEE Access, 11, pp. 95410-95431.
[3] Dubey S.R.,2021. A decade survey of content based image retrieval using deep learning. IEEE Transactions on Circuits and Systems for Video Technology,32(5), pp. 2687-2704.
[4] Liu Y., Ge Y., Wang F., Liu Q., Lei Y., Zhang D., andLu G., 2019. A rotation invariant HOG descriptor for tire pattern image classification. InICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 2412-2416.
[5] Kumar S., Singh M.K., andMishra M., 2023. Efficient deep feature based semantic image retrieval. Neural Processing Letters,55(3), pp. 2225-2248.
[6] Lowe D.G.,2004. Distinctive image features from scale-invariant keypoints. International Journal of Computer Vision,60(2), pp. 91-110.
[7] Dalal N., andTriggs B., 2005. Histograms of oriented gradients for human detection. In2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05), 1, pp. 886-893.
[8] Ojala T., Pietikainen M., andMaenpaa T., 2002. Multiresolution gray-scale and rotation invariant texture classification with local binary patterns. IEEE Transactions on Pattern Analysis and Machine Intelligence,24(7), pp. 971-987.
[9] Krizhevsky A., Sutskever I., andHinton G.E., 2017. ImageNet classification with deep convolutional neural networks. Communications of the ACM,60(6), pp. 84-90.
[10] Sharif Razavian A., Azizpour H., Sullivan J., andCarlsson S., 2014. CNN features off-the-shelf: an astounding baseline for recognition. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, pp. 806-813.
[11] Hendrycks D., andDietterich T., 2019. Benchmarking neural network robustness to common corruptions and perturbations.Arxiv Preprint Arxiv:1903.12261.
[12] Bayar B., andStamm M.C., 2017. On the robustness of constrained convolutional neural networks to jpeg post-compression for image resampling detection. In2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 2152-2156.
[13] Bilquees S., Dawood H., Dawood H., Majeed N., Javed A., andMahmood M.T., 2021. Noise resilient local gradient orientation for content‐based image retrieval.International Journal of Optics, 2021(1), 4151482.
[14] Goodfellow I.J., Pouget-Abadie J., Mirza M., Xu B., Warde-Farley D., Ozair S., Courville A., andBengio Y., 2014. Generative adversarial nets.Advances in Neural Information Processing Systems, 27.
[15] Ledig C., Theis L., Huszár F., Caballero J., Cunningham A., Acosta A., Aitken A., Tejani A., Totz J., Wang Z., andShi W., 2017. Photo-realistic single image super-resolution using a generative adversarial network. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 4681-4690.
[16] Kupyn O., Budzan V., Mykhailych M., Mishkin D., andMatas J., 2018. Deblurgan: blind motion deblurring using conditional adversarial networks. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 8183-8192.
[17] Zhang K., Zuo W., Chen Y., Meng D., andZhang L., 2017. Beyond a gaussian denoiser: residual learning of deep cnn for image denoising. IEEE Transactions on Image Processing,26(7), pp. 3142-3155.
[18] Vieira G.S., Fonseca A.U., Sousa N.M., Felix J.P., andSoares F., 2023. A novel content-based image retrieval system with feature descriptor integration and accuracy noise reduction.Expert Systems with Applications, 232, 120774.
[19] He K., Sun J., andTang X., 2010. Single image haze removal using dark channel prior. IEEE Transactions on Pattern Analysis and Machine Intelligence,33(12), pp. 2341-2353.
[20] Li B., Peng X., Wang Z., Xu J., andFeng D., 2017. Aod-net: all-in-one dehazing network. InProceedings of the IEEE International Conference on Computer Vision, pp. 4770-4778.
[21] Johnson J., Douze M., andJégou H., 2019. Billion-scale similarity search with GPUs. IEEE Transactions on Big Data,7(3), pp. 535-547.
[22] Ke G., Meng Q., Finley T., Wang T., Chen W., Ma W., Ye Q., andLiu T.Y., 2017. Lightgbm: A highly efficient gradient boosting decision tree.Advances in Neural Information Processing Systems, 30.
[23] Wang Z., Zheng Q., Lu J., andZhou J., 2020. Deep hashing with active pairwise supervision. InEuropean Conference on Computer Vision, pp. 522-538.
[24] Deng H., andCao Y., 2025. An adaptive fusion network integrating multi-source feature for remote sensing image-text retrieval. InProceedings of the 2025 3rd International Conference on Communication Networks and Machine Learning, pp. 103-109.
[25] Sun X., Jin L., andLi Z., 2019. Attention-aware feature pyramid ordinal hashing for image retrieval. InProceedings of the 1st ACM International Conference on Multimedia in Asia, pp. 1-6.
[26] Li F.F., Andreeto M., Ranzato M., andPerona P., 2022. Caltech 101 (1.0)[Data Set]. CaltechDATA[online].
[27] Yao B., Jiang X., Khosla A., Lin A.L., Guibas L.J., andFei-Fei L., 2011. Human action recognition by learning bases of action attributes and parts. InICCV, 2(6), 7.
[28] Wang J.Z., Li J., andWiederhold G., 2001. SIMPLIcity: semantics-sensitive integrated matching for picture libraries. IEEE Transactions on Pattern Analysis and Machine Intelligence,23(9), pp. 947-963.
[29] Malkov Y.A., andYashunin D.A., 2018. Efficient and robust approximate nearest neighbor search using hierarchical navigable small world graphs. IEEE Transactions on Pattern Analysis and Machine Intelligence,42(4), pp. 824-836.
[30] Vieira G.S., Fonseca A.U., andSoares F., 2023. CBIR-ANR: A content-based image retrieval with accuracy noise reduction.Software Impacts, 15, 100486.
[31] Zhao G., Zhang M., Liu J., andWen J.R., 2019. Unsupervised adversarial attacks on deep feature-based retrieval with GAN.Arxiv Preprint Arxiv:1907.05793.