Image Inpainting via Generative Multi-column with the aid of Deep Convolutional Neural Networks


Rajesh B, Assistant Professor, BASE University, Bengaluru and Muralidhara B L, Department of Computer Science & Application, Bangalore University, Bengaluru


BASE University Working Paper Series 07 /2021


Images can be described as visual representations or likeness of something (person or object or a scanned document) which can be reproduced or captured, e.g. a hand drawing, photographic material. The advent of the digital age has seen the rapid shift image storage technologies, from hard-copies to digitalized units in a less burdensome manner with the application of digital tools. The research aims to design a confidence-driven reconstruction loss while an implicit diversified Markov Random Field (MRF) regularization is adopted to enhance local details. The multi-column network combined with the reconstruction and MRF loss propagates local and global information derived from context to the target inpainting regions. Extensive experiments on challenging street view, face, natural objects and scenes manifest that our proposed method produces visual compelling results even without previously common post-processing. The research involves pre-trained Deep Convolutional Neural Network (DCNN) and their training networks like ResNet50, GoogleNet, AlexNet and VGG-16. The average PSNR performance of the proposed model is 24.64db and Structural Similarity Index Measure (SSIM) is 0.9018.
Keywords: Markov Random Field, Deep Convolutional Neural Network, ResNet50, GoogleNet, AlexNet, VGG-16, Structural Similarity Index Measure.