Volume - 4 | Issue - 3 | september 2022
DOI
10.36548/jiip.2022.3.006
Published
16 September, 2022
Colorization is not a guaranteed, but a feasible mapping between intensity and chrominance values. This paper presents a colorization system that draws inspiration from recent developments in deep learning and makes use of both locally and globally relevant data. One such property is the rarity of each color category on the quantized plane. The denoising model contains hybrid approach with cluster normalization through U-Net deep learning construction of framework. These are built on the basic U-Net design for segmentation. To eliminate gaussian noise in digital images, this article has developed and tested a generic deep learning denoising model. PSNR and MSE are used as performance measures for comparison purposes.
KeywordsU-Net noise removal colorization de-noising deep learning convolutional neural network