王林海，肖曉紅

?

基于改進(jìn)優(yōu)先權(quán)的Criminisi圖像修復(fù)算法的研究

*王林海1，肖曉紅2

（1.江蘇城鄉(xiāng)建設(shè)職業(yè)學(xué)院，江蘇，常州213147；2.井岡山大學(xué)電子與信息工程學(xué)院，江西，吉安 343009）

Crinimisi算法；圖像修復(fù)；可變大小樣本塊；改進(jìn)優(yōu)先權(quán)

1 概述

圖像修復(fù)技術(shù)是目前計(jì)算機(jī)的一種常用技術(shù)，其已被廣泛應(yīng)用于文物研究、新聞媒體以及藝術(shù)品修復(fù)等領(lǐng)域[1]。由于該技術(shù)可以最大限度地還原原有圖片的真實(shí)面貌，所以目前圖像修復(fù)技術(shù)的應(yīng)用范圍越來越廣泛，主要集中在文物保護(hù)領(lǐng)域、藝術(shù)品修補(bǔ)以及其他需要對(duì)一些原有圖片進(jìn)行修復(fù)的一些領(lǐng)域。隨著該技術(shù)的運(yùn)用范圍越來越廣泛，圖片修復(fù)技術(shù)的發(fā)展也越來越完善。

2　經(jīng)典的Criminisi算法分析

3　改進(jìn)的Criminisi算法

綜上所述，最終的改進(jìn)算法流程如下：

1）改進(jìn)優(yōu)先級(jí)的計(jì)算方法，引入相應(yīng)的調(diào)節(jié)參數(shù)，具體公式為

2）改進(jìn)匹配區(qū)域以及最佳匹配塊的確定；

3）置信值的改進(jìn)，增加一個(gè)特設(shè)定一顏色閾值T，進(jìn)行明確的區(qū)分。

4　實(shí)驗(yàn)結(jié)果與分析

圖1 大面積損壞下的不同算法的修復(fù)效果

圖2 不同算法的修復(fù)效果測(cè)試

圖3 不同技術(shù)的修復(fù)圖像SSIM值測(cè)試

5　結(jié)論

[1] 李光耀,耿瑞全,譚云蘭.基于區(qū)域分割和均方誤差改進(jìn)的圖像修復(fù)算法[J].井岡山大學(xué)學(xué)報(bào):自然科學(xué)版,2013,34(6):44-50.

[2] Jiao A S M, Tsang P W M, Poona T C.Restoration of digital off-axis Fresnel hologram by exemplar and search based image inpainting with enhanced computing speed[J].Computer Physics Communications, 2015, 193(8): 30-37.

[3] 耿衛(wèi)江.結(jié)構(gòu)相似度索引耦合最優(yōu)稀疏表示的大規(guī)模損壞圖像動(dòng)態(tài)修復(fù)[J].科學(xué)技術(shù)與工程,2014, 25(14): 107-114.

[4] Wang S, Guo W H, Huang T Z.Image inpainting using reproducing kernel Hilbert space and Heaviside functions[J].Journal of Computational and Applied Mathematics , 2017,311(2):551-564.

[5] Laurence C, Hussein F, Alain M.A Cahn-Hilliard System with a Fidelity Term for Color Image Inpainting[J]..Journal of Mathematical Imaging and Vision, 2016, 54(1): 117-131.

[6] Vahid K A, Farzin Y.Exemplar-based image inpainting using svd-based approximation matrix and multi-scale analysis[J].Multimedia Tools and Applications, 2017, 76(5): 7213-7234.

[7] Fakhry K.Textured image denoising using dominant neighborhood structure[J].Arabian Journal for Science and Engineering, 2014, 39(5):3759-3770.

[8] Mrinmoy G, Bhabatosh C.An image inpainting method using pLSA-based search space estimation [J].Machine Vision and Applications.2015,1(26):69-87.

[9] 林金勇,鄧德祥,顏佳.基于自適應(yīng)相似組稀疏表示的圖像修復(fù)算法[J].計(jì)算機(jī)應(yīng)用,2017,37(4):1169-1173.

[10] 王聰,楊晶,潘秀娟.基于光參量放大相位共軛特性的圖像修復(fù)與增強(qiáng)[J].物理學(xué)報(bào),2017,66(10):1-7.

[11] Li K S, Wei Y S, Yang Z.Image inpainting algorithm based on TV model and evolutionary algorithm[J].Soft Computing, 2016,20(3):885-893.

[12] 竇立云,徐丹,李杰.基于雙樹復(fù)小波的圖像修復(fù)[J].計(jì)算機(jī)科學(xué),2017,44(6A): 179-183.

Research of Image Inpainting Algorithm Based on Improved Priority Criminisi method

*WANG Lin-hai1；XIAO Xiao-hong2

（1.Jiangsu Urban and Rural Construction College, Changzhou,Jiangsu 213147, China；2.School of Electronics and Information Engineering, Jinggangshan University, Ji’an, Jiangxi 343009, China）

Aiming at overcoming the shortcomings of inaccurate priority calculation in Criminisi algorithm and the problem of sample block solidification size, this paper proposes an improved priority-based Criminisi image restoration algorithm.In the calculation of priority function, the weighted sum of data item, confidence item and neighborhood correlation item is calculated to optimize the priority of Crinisi algorithm, and the continuity of calculation function is effectively enhanced to ensure the continuity of the repaired image structure.Variable block size is used to enhance local coordination and boundary smoothness, improve the adaptability of Criminisi image restoration algorithm to different smoothness of the image, and ameliorate the restoration performance of the algorithm.Experimental results show that the proposed algorithm has good visual effects.

Criminisi algorithm; image inpainting; variable size block; improved priority

TP391.41

A

10.3969/j.issn.1674-8085.2018.06.008

2018-08-03；

2018-10-04

*王林海(1982-)，男，江蘇常州人，講師，碩士，主要從事計(jì)算機(jī)技術(shù)研究(E-mail:gy798541@163.com);

肖曉紅(1971-)，女，江西吉安人，副教授，碩士，主要從事圖像處理、智慧城市等研究(E-mail: jgsxxh5659@126.com).

1674-8085(2018)06-0044-05