OPTIMIZATION ASSISTED AUTOREGRESSIVE METHOD WITH DEEP CONVOLUTIONAL NEURAL NETWORK-BASED ENTROPY FILTER FOR IMAGE DEMOSAICING
Abstract
The natural scenes are captured by digital cameras that adopt a single charged-coupled device (CCD) or Complementary Metal Oxide Semiconductor (CMOS) sensor with a Color Filter Array (CFA). To reconstruct a full color image from the mosaiced CFA image, the missing components of color are recovered by a technique called color demosaicing. This research work introduces Deep Convolution Neural Network (DCNN)-based optimization assisted autoregressive method for image demosaicing. Here, the proposed optimization algorithm, termed Adaptive Autoregressive Water Wave Optimization algorithm (Adaptive Autoregressive-WWO), that is formed by combining Conditional autoregressive value at risk (CAViaR) model and Water Wave Optimization algorithm (WWO). Fusion process is carried out for residual images to generate the final demosaiced image based on entropy measure. Here, the output generated from DCNN is the first residual image. The LPA-ICI filter utilizes the optimization algorithm that generated second order polynomial coefficients to produce the second residual output image. Moreover, this proposed method is evaluated for its performance using metrics, such as Peak signal-to-noise ratio (PSNR) and Second Derivative like Measurement (SDME) and attained the highest PSNR values of 40.379dB and highest SDME values of 50.675dB.

Authors
C. Anitha Mary, A. Boyed Wesley
Nesamony Memorial Christian College, India

Keywords
Imagedemosaicing, Local Polynomial Approximation and Intersection of Confidence Interval Filter, Entropy, Fusion Process, Deep Convolution Neural Network
Yearly Full Views
JanuaryFebruaryMarchAprilMayJuneJulyAugustSeptemberOctoberNovemberDecember
000000010000
Published By :
ICTACT
Published In :
ICTACT Journal on Soft Computing
( Volume: 13 , Issue: 3 , Pages: 2977 - 2985 )
Date of Publication :
April 2023
Page Views :
210
Full Text Views :
2

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.