A Method for Estimating the Parameters of a Linearly Blurred Image Based on Comparison with Two-dimensional Histograms of Brightness Gradients of an Artificially Blurred Standard Image
( Pp. 24-32)

Task. Image blur is one of the most common defects in photography. In the absence of information about the optical system at the time of shooting, it is impossible to accurately determine the blur model and its parameters. For small images, neural network approaches or the use of mathematical methods based on the Radon, Hough transform and the cepstral method are possible. However, in the case of large images, these methods are not applicable due to the high computational complexity, so there is a need to develop a new method for processing such images. Model. A new method for estimating linear blur parameters on uniformly distorted images is proposed, based on comparing two-dimensional histograms of gradients with pre-calculated histograms obtained from a reference image with various simulated blur parameters. Mostly this method is statistical, using classical mathematical methods of image processing. Results. The proposed method shows qualitative results for any linear blur parameters. At the same time, the closer the reference image is to the original one in the gradient histogram, the more accurate the result is, up to zero error. If we take as a reference image not the closest one, but from the same class of images, the error as a result will be no more than half a pixel in each of the directions of the blur. Practical significance. The proposed method is applicable for processing high-resolution images with linear blur. The developed algorithm is applicable, for example, for processing satellite images. Value. The proposed method has a great advantage over known mathematical methods and neural network methods for determining lubrication parameters due to its high accuracy with low computational complexity. Its application will bring significant benefits in real-time image processing.

Kharlamov S.G. and Fedorov V.B. A method for estimating the parameters of a linearly blurred image based on comparison with two-dimensional histograms of brightness gradients of an artificially blurred standard image. Computational Nanotechnology. 13, 1 (2026), 24–32. DOI: 10.33693/2313-223X-2026-13-1-24-32. EDN: LYTNIZ

Gulyaev Yu.V. et al. Correction of the spatial spectrum distorted by an optical system using the reference image method. Part 3. Universal reference spectrum. Journal of Radio Electronics. 2013. No. 12. (In Rus.). EDN: SBNUVB.
Fedorov V.B., Kharlamov S.G., Fedorov A.V. Image reconstruction using a discrete point spread function obtained taking into account the finiteness of the pixel size. Russian Technological Journal. 2025. Vol. 13. No. 2. Pp. 143–154. (In Rus.). DOI: 10.32362/2500-316X-2025-13-2-143-154. EDN: GXAGAW.
Asai H., Oyamada Y., Pilet J., Saito H. Cepstral analysis based blind deconvolution for motion blur. In: Proceedings of the IEEE International Conference on Image Processing (ICIP). Hong Kong: IEEE, 2010. Pp. 1153–1156. DOI: 10.1109/ICIP.2010.5651299.
Carbajal G., Vitoria P., Lezama J., Musé P. Blind motion deblurring with Pixel Wise Kernel Estimation via Kernel Prediction networks. arXiv preprint. 2023. arXiv: 2308.02947.
Cho T.S., Paris S., Horn B.K.P., Freeman W.T. Blur kernel estimation using the Radon Transform. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Colorado Springs: IEEE, 2011. Pp. 241–248. DOI: 10.1109/CVPR.2011.5995479.
Debarnot V., Weiss P. Deep-blur: Blind identification and deblurring with convolutional neural networks. Biological Imaging. 2024. Vol. 4. P. e13. DOI: 10.1017/S2633903X24000096.
Deshpande A.M., Patnaik S. A novel modified cepstral based technique for blind estimation of motion blur. Optik. 2014. Vol. 125. No. 2. Pp. 606–615. DOI: 10.1016/j.ijleo.2013.05.189.
Iraei I., Sharifi M., Baleghi Y. A new approach to enhance the estimation of blur parameters in blurred images. Optik. 2020. Vol. 224. P. 165298. DOI: 10.1016/j.ijleo.2020.165298.
Jha T.N. Velocity detection from a motion blur image using Radon Transformation. Tribhuvan University Journal. 2018. Vol. 32. No. 2. Pp. 243–248. DOI: 10.3126/tuj.v32i2.24721.
Mittal A., Soundararajan R., Bovik A.C. Making a “completely blind” image quality analyzer. IEEE Signal Processing Letters. 2013. Vol. 20. No. 3. Pp. 209–212. DOI: 10.1109/LSP.2012.2227726.
Pan J., Ren W., Hu Z., Yang M.-H. Learning to deblur images with exemplars. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2019. Vol. 41. No. 6. Pp. 1412–1425. arXiv: 1805.05503. DOI: 10.1109/TPAMI.2018.2832125.
Varela L.G. et al. Estimation of motion blur kernel parameters using regression convolutional neural networks. Journal of Electronic Imaging. 2024. Vol. 33. No. 2. P. 023062. DOI: 10.1117/1.JEI.33.2.023062.
Wang T., Shu H., Jia H., Li B. Blind image quality assessment using natural scene statistics in the gradient domain. In: Proceedings of the 8th Asia Modelling Symposium (AMS). 2014. Pp. 56–60. DOI: 10.1109/AMS.2014.22.
Wang Y. et al. Improved scheme of estimating motion blur parameters for image restoration. Digital Signal Processing. 2017. Vol. 65. Pp. 11–18. DOI: 10.1016/j.dsp.2017.02.010.
Zou H., Suganuma M., Okatani T. Reference based motion blur removal: Learning to utilize sharpness in the reference image. CoRR. 2023. arXiv: 2307.02875.

gradient histograms, linear blur, standard image, blind deconvolution, distribution comparison, distance metric between histograms.