【正文】
ns and divisions in Galois Fields are plex operations and, since decoding calculation steps require a large amount of multiplications, very much care has been given to the i mplementation of GF inner products. Resorting to two dedicated lookup tables has permitted to achieve very good performance with reduced plexity. Furthermore, this RS implementation has excellent figures of modularity, since the five calculation steps can be puted with the same core. Selection of proper input data and signals is a trivial task, demanded to RS core control unit. 4 Results In the following we present the results of tests per formed on the proposed system. Results on visual quality witness the performance improvement that can be obtained by protecting the JPEG2020 stream with ULP encoding. Moreover, profiling results of the DSP /FPGA system are reported. Visual quality We report here the system performance in terms of the quality of service delivered at 中北大學(xué) 2020 屆畢業(yè)設(shè)計(jì)說(shuō)明書(shū) 第 8 頁(yè) 共 15 頁(yè) the receiver. The adopted joint sourcechannel coding strategy allows to couple the powerful pression capabilities of JPEG2020 with the error correction provided by RS codes. We tested the proposed approach using images of 256x256 size, transmission rate equal to bpp, . 87 ATM packets of 47 bytes (plus 1 byte devoted to sequence numbering)。因此,圖片傳送技術(shù)必須要得到其他技術(shù)的支撐,如圖像恢復(fù)技術(shù),或者至少是能夠減小傳送錯(cuò)誤影響的技術(shù),已經(jīng)制定的JPEG2020圖片壓縮標(biāo)準(zhǔn)與上述要求完全相吻合,并且包含了一部分錯(cuò)誤檢索和修復(fù)工具。 JPEG2020的內(nèi)部結(jié)構(gòu)是基于代碼轉(zhuǎn)化端口創(chuàng)建。由此可將最重要的信息放在字符串的開(kāi)始。在譯碼器內(nèi),由于 RS代碼的錯(cuò)誤修正容量, ITH行能被精確恢復(fù),因此文件刪除數(shù)量比 TI少。然而,許多標(biāo)準(zhǔn)任有可能擴(kuò)展,一個(gè) DSP程序的執(zhí)行將提高圖像吻合的等級(jí)。最后,包含在字符串中的每個(gè)文件包均被讀出,代碼塊的數(shù) 據(jù)和范圍信息也被提取,這些都將輸入到 EBCOT譯碼器中。從形式上講, EBCOT由三個(gè)主要階段組成, SP、 MR、 CL上述每一個(gè)步驟采用四個(gè)原始的譯碼步驟,稱(chēng)為 Zero Coding, Sign Coding, Magnitude Refinement Coding, and Run Length Coding,訪問(wèn)命令遵循 SPMRCL的次序;值得注意的是,給定代碼模塊的每一樣本盡在上訴三個(gè)步驟之一中運(yùn)行,就計(jì)算機(jī)器件的復(fù)雜程度所受關(guān)注而言, CL在破譯最有意義的是要求最大速率。 JPEG2020運(yùn)行遵循先行后列的濾波順序。前者收集從網(wǎng)絡(luò)上接收到的文件包,按列填充到矩陣,并且逐行轉(zhuǎn)化到 RS譯碼器的內(nèi)核。 4 設(shè)計(jì) 結(jié)果 接下來(lái)我們將看到建議系統(tǒng)的測(cè)試運(yùn)行結(jié)果。該圖片用未加 FEC保護(hù)的 JPEG2020實(shí)現(xiàn),與僅使用單獨(dú)擦除文件包的左圖成鮮明對(duì)比;質(zhì)量的優(yōu)劣是顯而易見(jiàn)的。這樣就開(kāi)辟了一條低難度,純軟件多媒體系統(tǒng)的應(yīng)用實(shí)施新道路,它基于 幀 幀 MotionJPEG2020 內(nèi)核,完成視頻在通信系統(tǒng)中的多進(jìn)程傳送。結(jié)果通過(guò)用與 Sect. 同的狀況來(lái)實(shí)現(xiàn), 還有使用 (9, 7)濾波器的五個(gè)轉(zhuǎn)化層和尺寸為 8*8的代碼塊。采用聯(lián)合解碼戰(zhàn)略允許將 JPEG2020強(qiáng)勁的壓縮容量和 RS代碼提供的錯(cuò)誤修正組合在一起。值得注意的是,乘和除在GF中是復(fù)雜操作,并且由于解碼運(yùn)算步驟中有大量的乘法運(yùn)算,所以在執(zhí)行 GF內(nèi)部乘積運(yùn)算操作時(shí)要特別小心。為了充分利用 0附近區(qū)域的動(dòng)態(tài)變化,筆者預(yù)測(cè)到了 DCT 技術(shù),該技術(shù) 作為 DC的重要組成部分,能使低通系數(shù)變化范圍的巨幅擴(kuò)大。 統(tǒng)一標(biāo)量 規(guī)模 根據(jù), JPEG2020支持下稱(chēng)為標(biāo)量統(tǒng)一。為破譯一代碼模塊, EBCOT總是從最有意義的開(kāi)始,然后逐步轉(zhuǎn)移到最不重要的去執(zhí)行。 JPEG2020解碼器模塊組件全部的用到了 DSP中,主要有四大部分組成 :語(yǔ)法分析 , EBCOT統(tǒng)一建制和波的逆變換?;?RS代碼技術(shù)的附加字符串保護(hù)附件,已從 XILINX被運(yùn)用到了 Virtex XCVl000 FPGA器件,該附件能實(shí)現(xiàn) 與 NSP同時(shí)交換數(shù)據(jù)信息。另一方面,文件包丟失經(jīng)常發(fā)生在有潛在堵塞危險(xiǎn)的網(wǎng)絡(luò)中,實(shí)際上往往 是文件丟失。前者基于字符段端口技術(shù)(最有意義的字符頻段首先被傳送)和環(huán)境系統(tǒng)模型化技術(shù)以及算數(shù)代碼技術(shù)。 JPEG2020圖像壓縮技術(shù) 中北大學(xué) 2020 屆畢業(yè)設(shè)計(jì)說(shuō)明書(shū) 第 11 頁(yè) 共 15 頁(yè) JPEG2020是針對(duì)與靜止圖像編碼的全新 ISO標(biāo)準(zhǔn),并且不斷致力于 提供緊跟多媒體技術(shù)新潮流的新標(biāo)準(zhǔn)。該系統(tǒng)的執(zhí)行正是基于融合的 DSP和 FPGA技術(shù)構(gòu)建,它允許同時(shí)運(yùn)行多個(gè)計(jì)算機(jī)程序,這樣就使得系統(tǒng) 可以實(shí)現(xiàn)高效率運(yùn)行。 this is sketched in the righthandside box of Fig. 1. Adaptive ReedSolomon packet protection Although JPEG2020 embodies advanced error concealment techniques to mitigate the effect of errors, it does neither contain, nor specify any error correction method, in order to recover lost packets. On the other hand, packet losses are likely to occur in a work potentially subject to congestion, as is often the case in practice. In order to overe this problem, a technique has been recently proposed, called Unequal Loss Protection (ULP), and based on the joint use of RS codes and packet interleaving, as shown in the lefthandside of Fig. 1. Let us consider a maximum rate allocated to the image transmission, . N packets of size L。 then, the transform coefficients are quantized, independently for each subband, with an embedded deadzone quantizer. Each subband of the wavelet deposition is divided into rectangular blocks (codeblocks), which are in dependently encoded with EBCOT (Embedded Block Coding with Optimized Truncation) 。 RS code allocation optimized for an exponential packet loss model with 20% mean loss rate. In Fig. 2, visual results are presented. The righthandside image has been obtained by the proposed demonstrator, using RS codes and JPEG2020, with 20% lost packets. The image achieved by using JPEG2020 without FEC protection, and a single erased packet, is shown on the lefthandside for parison。 本文 闡述了用 JPEG2020技術(shù)對(duì)圖片傳輸過(guò)程中易丟失文件的網(wǎng)絡(luò)實(shí)施檢錯(cuò)糾錯(cuò)的過(guò)程。圖像信息可被分為幾個(gè)子圖信息,以減小所需內(nèi)存,降低對(duì)計(jì)算的要求。JPEG2020幾乎使用了與之相同的步驟(除了速率分配器采用相反的順序):句法分析,用 EBCOT對(duì)代碼模塊解碼,數(shù)目取反, NWT取反,子圖拼湊。 中北大學(xué) 2020 屆畢業(yè)設(shè)計(jì)說(shuō)明書(shū) 第 12 頁(yè) 共 15 頁(yè) 圖 1 系統(tǒng)的體系結(jié)構(gòu) 譯碼制度 的擬定 系本文中給我們展示了圖片在有文件包丟失的網(wǎng)絡(luò)傳送中的全譯碼器的示范者。另一方面, FPGA與 Reed Solomon的核心吻合的很好,需要將交流系統(tǒng)的錯(cuò)誤修復(fù)植入。 EBCOT 緊接著字符串的句法分析