IRGUN: Improved Residue based Gradual Up-Scaling Network for Single Image Super Resolution

Sharma, M and Mukhopadhyay, R and Upadhyay, A and Shukla, A and Koundinya, SA and Chaudhury, S (2018) IRGUN: Improved Residue based Gradual Up-Scaling Network for Single Image Super Resolution. In: Conference on Computer Vision and Pattern Recognition (CVPR) , June 18-22, 2018, Salt Lake City . (Submitted)

Preview

PDF - Submitted Version Download (7Mb) | Preview

Abstract

Convolutional neural network based architectures hm•e achieved decent perceptual quality super resolution on nat­ ural imagesfor small scaling factors (2X and 4X). Howe1•e1; image super-resolution for large magnication factors ( 8X) is an extremely challenging problem for the computer l'i­ sion community. In this paper, we propose a novel Improved Residual based Gradual Up-Scaling Network (IRGUN) to improve the quality of the super-resolved image for a large magnification factor. IRGUN has a Gradual Upsampling and Residue-based Enhancment Network (GUREN) which comprises of series of Up-scaling and Enhancement hlocks (UEB) connected end-to-end and fine-tuned together to give a gradual magnification and enhancement. Due to the per­ ceptual importance of the luminance in super-resolution, the model is trained on luminance (Y) channel of the YCbCr image. Whereas, the chrominance components (Cb and Cr) channel are up-scaled using bicubic interpolation and com­ bined with super-resolved Y channel of the image. which is then converted to RGB. A cascaded JD-RED architec­ ture trained on RGB images is utilized to incorporate its inter-channel correlation. In addition to this, the training methodology is also presented in the paper. In the train­ ing procedure, the weights of the previous UEB are used in the next immediate UEB for faster and better convergence. Each UEB is trained on its respective scale hy taking the output image of the previous UEB as input and correspond­ ing HR image of the same scale as ground truth to the suc­ cessive UEB. All the UEBs are then connected end-to-end and.fine tuned. The IRGUN recoversfine details effectively at large (8X) magnification factors. The efficiency of JR­ CUN is presented on various benchmark datasets and at different magnification scales.

Item Type:	Conference or Workshop Item (Paper)
Subjects:	Electronic Systems > Digital Systems
Divisions:	Electronic Systems
Depositing User:	Mr. Rabin Chatterjee
Date Deposited:	27 Jul 2021 09:39
Last Modified:	27 Jul 2021 09:39
URI:	http://ceeri.csircentral.net/id/eprint/376

Actions (login required)

View Item