Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/7456
DC FieldValueLanguage
dc.contributor.advisor歐陽彥杰zh_TW
dc.contributor.advisorYen-Chieh Ouyangen_US
dc.contributor.author許進裕zh_TW
dc.contributor.authorSheu, Jinn-Yuhen_US
dc.date2001zh_TW
dc.date.accessioned2014-06-06T06:40:04Z-
dc.date.available2014-06-06T06:40:04Z-
dc.identifier.urihttp://hdl.handle.net/11455/7456-
dc.description.abstract我們提出了一個運用於連續影像壓縮中新的預測性運動向量搜尋演算法,並且將之取名為新十字搜尋演算法,此種演算法可以在維持影像品質的前提下有效的提升運動向量的尋找速度。為了降低搜尋過程中陷入區域極小值內的機率,我們採取了多重搜尋起點的策略並且搭配對區塊匹配中像素的次取樣機制來維持搜尋的速度。而在計算區塊誤差時,我們允許在誤差值已高於最小誤差時,中途停止此候選運動向量的區塊誤差計算,以避免計算資源的浪費。此外,當目前所計算出的最小區塊誤差低於所設定之門檻值時,我們也會提早結束此運動向量的尋找,並且以此最小誤差區塊的位移作為該區愧的運動向量。模擬結果顯示使用本演算法相對於預測性找尋演算法最多可減少百分之八十八的計算量,而平均也可減少百分之七十八的計算量,若與向量找尋演算法相比,則本演算法有效的減少了許多不必要的運算並且提升了影像品質。透過以上的模擬結果,我們知道新十字搜尋演算法的確能在維持影像品質的前提下,有效的加快運動向量的搜尋速度,因此,本演算法可以提供作為即時影像壓縮編碼器中運動向量搜尋演算法的新選擇。zh_TW
dc.description.abstractA new cross search (NCS) algorithm, which can efficiently reduce the computational load, is proposed. To reduce the probability of being trapped into local minimum in motion vector search, the NCS algorithm initiates its search with multiple starting points meanwhile utilize the pixel sub-sampling method to maintain a low computational load. In order to early reject non-possible candidate motion vector, the NCS algorithm adopts a halfway-stop mechanism into the computation process of block distortion measurement. Furthermore, we introduce a threshold value to early stop if the current minimum block error is less than the threshold value. Simulation results show that by using the NCS algorithm we can reduce up to 88% computational load and 78% on average compared to the PSA meanwhile maintaining similar PSNR and MSE performance. When compares with VSA, the NCS can reduce redundant computation and improve the performance. The proposed NCS algorithm is suitable for real-time video encoding as it can speed up the encoder without sacrificing PSNR performance.en_US
dc.description.tableofcontentsChapter 1 1 Introduction1 1.1Motivation1 1.2Overview of Video Compression2 1.2.1Lossless and lossy compression3 1.2.2Transformation of Color Space4 1.2.3Motion Estimation and Compensation5 1.3Organization of Thesis6 Chapter 2 8 Architecture of Block-based Video Compression8 2.1Introduction8 2.2System Architecture of H.263 Encoder9 2.3Matching Criteria11 2.4Summary12 Chapter 3 13 Overview on Previously Proposed Motion Estimation Algorithms13 3.1Introduction13 3.2Full Search Algorithm14 3.3Logarithmic Step Search Algorithms15 3.4Predictive Search Algorithms20 3.4.1Correlation in Motion Vectors20 3.4.2Concept of Motion Vector Prediction22 3.4.3Predictive Search Algorithm (PSA)23 3.4.4Vector Search Algorithm (VSA)25 3.5Summary27 Chapter 4 28 Proposed New Cross Search Algorithm28 4.1Overview28 4.2Multiple Starting Points29 4.3The Sub-sampling Pattern Comparison for Block-matching31 4.4Half-way Stop Mechanism38 4.5The New Cross Search (NCS) Algorithm39 4.6Simulation Results43 4.6.1Computational Load Comparison44 4.6.2Performance Comparison48 Chapter 5 55 Conclusion and Future Works55 5.1Conclusion55 5.2Future Works57 Bibliography58en_US
dc.language.isoen_USzh_TW
dc.publisher電機工程學系zh_TW
dc.subjectNew cross search algorithmen_US
dc.subject新十字搜尋演算法zh_TW
dc.subjectmotion estimationen_US
dc.subjectpredictive search algorithmen_US
dc.subjecthalfway-stop mechanismen_US
dc.subjectmultiple starting pointsen_US
dc.subject運動估測zh_TW
dc.subject預測性搜尋演算法zh_TW
dc.subject中途停止機制zh_TW
dc.subject多重搜尋起點zh_TW
dc.title一種以區塊為基礎的預測性運動向量估測演算法zh_TW
dc.titleA Modified Predictive Algorithm for Block-based Motion Estimationen_US
dc.typeThesis and Dissertationzh_TW
item.openairecristypehttp://purl.org/coar/resource_type/c_18cf-
item.openairetypeThesis and Dissertation-
item.cerifentitytypePublications-
item.fulltextno fulltext-
item.languageiso639-1en_US-
item.grantfulltextnone-
Appears in Collections:電機工程學系所
Show simple item record
 
TAIR Related Article

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.