基于DCT的HEVC幀內(nèi)預(yù)測快速算法研究及優(yōu)化
發(fā)布時間：2018-11-28 12:16
【摘要】：High Efficiency Video Coding(HEVC)是H.264/AVC之后的新一代視頻編碼標準,其目的是為了達到更高的編碼效率,更好的網(wǎng)絡(luò)適應(yīng)性,滿足人們對高質(zhì)量視頻的需求。HEVC保留了預(yù)測加變換的混合編碼框架,同時改進和采納了諸多近年來視頻編碼的研究成果。在幀內(nèi)預(yù)測方面采用了靈活的編碼結(jié)構(gòu)和多角度預(yù)測,包括編碼單元(CU)、預(yù)測單元(PU)、變換單元(TU)和35種預(yù)測模式。然而,這些新技術(shù)的引入在提高編碼性能的同時提高了編碼復(fù)雜度。 為了解決HEVC編碼器中運算量龐大的問題,論文研究和改進了兩種幀內(nèi)預(yù)測快速算法。第一,針對CU的四叉樹劃分,研究和改進了基于離散余弦變換(Discrete Cosine Transform, DCT)的幀內(nèi)預(yù)測CU快速深度選擇算法,通過統(tǒng)計和分析DCT系數(shù)與CU紋理復(fù)雜度的關(guān)系,采用DCT的低頻交流系數(shù)并結(jié)合率失真代價進行CU紋理復(fù)雜度的判定,從而提前結(jié)束CU的劃分。第二,針對35種幀內(nèi)預(yù)測模式,論文研究和改進的算法利用DCT預(yù)測編碼塊的紋理方向,減少進行粗選擇的模式數(shù),并且通過分析粗選擇后預(yù)測模式的特點,采用分組算法,減少最后進行率失真代價計算的模式數(shù)。 實驗結(jié)果表明,論文研究的兩種算法在官方測試模型HM10.0,全I幀的條件下,平均縮短30.14%和24.88%的編碼時間,碼率和圖像質(zhì)量幾乎保持不變。
[Abstract]:High Efficiency Video Coding (HEVC) is a new generation of video coding standard after H.264/AVC. Its purpose is to achieve higher coding efficiency and better network adaptability. To meet the needs of high quality video, HEVC retains the hybrid coding framework of predictive plus transform, while improving and adopting many recent research achievements in video coding. In intra-frame prediction, flexible coding structure and multi-angle prediction are adopted, including coding unit (CU), prediction unit (PU), transform unit (TU) and 35 prediction modes. However, the introduction of these new techniques not only improves the coding performance, but also increases the coding complexity. In order to solve the problem of huge computation in HEVC encoder, two fast intra prediction algorithms are studied and improved in this paper. Firstly, aiming at the quadtree partition of CU, the fast depth selection algorithm of intra prediction CU based on discrete cosine transform (Discrete Cosine Transform, DCT) is studied and improved. The relationship between DCT coefficients and CU texture complexity is analyzed by statistics and analysis. The low frequency AC coefficient of DCT and the cost of rate-distortion are used to judge the complexity of CU texture, so that the division of CU is finished in advance. Secondly, aiming at 35 intra prediction modes, this paper uses DCT to predict the texture direction of coding block, and reduces the number of rough selection modes. By analyzing the characteristics of prediction mode after rough selection, the paper adopts grouping algorithm. Reduce the number of modes for final rate-distortion cost calculations. The experimental results show that the two algorithms in this paper can reduce the coding time by 30.14% and 24.88% under the condition of the official test model HM10.0, full I frame, and the code rate and image quality are almost unchanged.
【學位授予單位】：南京理工大學
【學位級別】：碩士
【學位授予年份】：2014
【分類號】：TN919.81
本文編號：2362796
[Abstract]:High Efficiency Video Coding (HEVC) is a new generation of video coding standard after H.264/AVC. Its purpose is to achieve higher coding efficiency and better network adaptability. To meet the needs of high quality video, HEVC retains the hybrid coding framework of predictive plus transform, while improving and adopting many recent research achievements in video coding. In intra-frame prediction, flexible coding structure and multi-angle prediction are adopted, including coding unit (CU), prediction unit (PU), transform unit (TU) and 35 prediction modes. However, the introduction of these new techniques not only improves the coding performance, but also increases the coding complexity. In order to solve the problem of huge computation in HEVC encoder, two fast intra prediction algorithms are studied and improved in this paper. Firstly, aiming at the quadtree partition of CU, the fast depth selection algorithm of intra prediction CU based on discrete cosine transform (Discrete Cosine Transform, DCT) is studied and improved. The relationship between DCT coefficients and CU texture complexity is analyzed by statistics and analysis. The low frequency AC coefficient of DCT and the cost of rate-distortion are used to judge the complexity of CU texture, so that the division of CU is finished in advance. Secondly, aiming at 35 intra prediction modes, this paper uses DCT to predict the texture direction of coding block, and reduces the number of rough selection modes. By analyzing the characteristics of prediction mode after rough selection, the paper adopts grouping algorithm. Reduce the number of modes for final rate-distortion cost calculations. The experimental results show that the two algorithms in this paper can reduce the coding time by 30.14% and 24.88% under the condition of the official test model HM10.0, full I frame, and the code rate and image quality are almost unchanged.
【學位授予單位】：南京理工大學
【學位級別】：碩士
【學位授予年份】：2014
【分類號】：TN919.81
【參考文獻】
相關(guān)期刊論文 前5條
1 黃鐵軍,高文;AVS標準制定背景與知識產(chǎn)權(quán)狀況[J];電視技術(shù);2005年07期
2 石飛宇;劉昱;;一種HEVC快速幀內(nèi)模式判斷算法[J];電視技術(shù);2013年11期
3 趙文強;沈禮權(quán);張兆楊;;HEVC幀內(nèi)預(yù)測算法的優(yōu)化[J];電視技術(shù);2012年08期
4 張曉蕓,宋建社,鄭永安;視頻編碼技術(shù)研究進展[J];計算機工程;2004年14期
5 蔣潔;郭寶龍;莫瑋;焦昶哲;孫偉;;利用平滑區(qū)域檢測的HEVC幀內(nèi)編碼快速算法[J];西安電子科技大學學報;2013年03期
，本文編號：2362796
本文鏈接：http://sikaile.net/kejilunwen/wltx/2362796.html
最近更新
教材專著
