Identification of Relevant Images from Ancient Paintings using Statistical Distribution

G G Naidu,Y Srinivas

doi:10.17485/ijst/2018/v11i17/117688

Abstract

Objectives: This article highlights a novel concept of identifying the ancient images along with the artists by proposing a model based on a Generalized Bivariate Laplacian Mixture Model (GBLMM) approach for this purpose. Methods/Statistical Analysis: Conservative Chinese paintings emulate the exquisiteness of Chinese sculpture. Most of these descriptions are digitized and are available in the internet. One of challenging task is to retrieve the images of significance more efficiently and proficiently. This article highlights a pioneering idea to spot these ancient images along with the artist by proposing a model based on a Generalized Bivariate Laplacian Mixture Model (GBLMM) approach. Findings: The methodology is subjected to the application of ancient Chinese paintings. The results derived are evaluated against quality metrics such as image fidelity, peak signal noise ratio, structured coefficients, average difference and mean squared error. The proposed model accomplished over 97% of classification accuracy over a dataset containing 3750 traditional Chinese paintings. Application/Improvements: In order to ascertain the quality of segmentation, metrics like Image Fidelity (IF), Mean Squared Error (MSE) and Peak Signal Noise Ratio (PSNR) are considered. This model can be very much useful for the archeologists to identify the ancient paintings. Keywords: Ancient Paintings, GBLMM, Mixture Model, Quality Metrics, Retrieval

Highlights

Image processing aims at development of algorithms for better understanding of the image details, ability to analyze the details, enhancing the images and retrieval of significant regions of interest from these images
The Meta tags and the color of the image are taken as features. These features are given as inputs to the Generalized Bivariate Laplacian Mixture Model (GBLMM) for the retrieval of relevant images and the identification of the artist together with the period of painting
The main intuition to identify the specks within every image region, GBLMM is considered

Summary

Introduction

Image processing aims at development of algorithms for better understanding of the image details, ability to analyze the details, enhancing the images and retrieval of significant regions of interest from these images. Identification of Relevant Images from Ancient Paintings using Statistical Distribution above presented methodologies, classifications of images based on the attributes are mostly preferred. Knowledge discovery of artistic influences focused on the comparison between the painted images, with the visual similarity[14] This assessment is based on computer vision approaches and machine learning. Image processing, image retrieval, machine learning and artificial intelligence They highlighted two different clustering algorithms namely, hierarchical and k-Means. The Meta tags and the color of the image are taken as features These features are given as inputs to the Generalized Bivariate Laplacian Mixture Model (GBLMM) for the retrieval of relevant images and the identification of the artist together with the period of painting. The results derived together with performance evaluation models are summarized in the Section 6

Generalized Bivariate Laplacian Mixture Model

Major Challenges

Dataset Considered

Features Selection and Methodology

Conclusions