【導(dǎo)讀】常用的圖像處理方法有圖。像增強、復(fù)原、編碼、壓縮等。才處于高頻區(qū)域中。因為在圖像的數(shù)字化和傳輸中常有噪聲出現(xiàn)，而這部分干擾信息主要集中在。應(yīng)該是既能消去噪聲對圖像的影響又不使圖像細(xì)節(jié)變模糊。信息，必須對圖像進行去噪預(yù)處理。整理文獻(xiàn)，研究現(xiàn)有基于小波變換的圖像去噪算法，嘗試對現(xiàn)有算法做出改進；在MATLAB下仿真驗證基于小波變換的圖像去噪算法。理，方法，結(jié)論”的要素，對所研究內(nèi)容作出詳細(xì)有條理的闡述。1-3周：查找資料，文獻(xiàn)。12-14周：分析試驗結(jié)果，對比各種算法的優(yōu)點和缺點，嘗試改進算法。15-17周：撰寫畢業(yè)論文，完成畢業(yè)答辯。圖像去噪領(lǐng)域受到越來越多的關(guān)注。方法和基于小波變換的濾波去噪方法，其基本思想是先對含噪圖像進行小波變換，再對高頻系數(shù)進行閾值去噪或濾。波去噪處理，最后進行小波反變換，實現(xiàn)基于小波的圖像去噪。

　　

【正文】 s therein) and is applied in many existing munication systems. However, only few results are available for noncoherent equalization schemes, . , for the bination of linear or nonlinear equalization and noncoherent detection. Such noncoherent receivers have the important advantage offset than coherent equalization schemes. Noncoherent linear minimum mean squared error ( MMSE ) equalization has been proposed in [3, 4] , while noncoherent decision feedback equalization ( DFE ) has been regarded in [5, 6] . These noncoherent equalizers have in mon that they are only designed for MDPSK(Mary differential phaseshift keying), which may be considered as a special case of MDAPSK. However, recently MDAPSK ( Mary differential amplitude/phaseshift keying) constellations with more than one amplitude level have bee very popular because of amplitude level have bee very popular because of their high spectral efficiency (see . [7, 8] and references therein). Therefore, it is desirable to derive robust noncoherent linear equalization ( NLE ) schemes for general MDAPSK ( Mary differential amplitude/phaseshift keying ) signals. A 陜西理工學(xué)院畢業(yè)設(shè)計 第 16 頁 共 45頁 noncoherent MMSE equalizer for 16DAPSK has been considered in [9]。 however, it turned out that a theoretical analysis of this equalizer is very difficult if impossible. Similar to the schemes giver in [4, 9], the proposed NLE ( noncoherent linear equalization ) scheme consists of a linear equalizer bined with a decision feedback differential detector (DF DD) [7, 8]. In contrast to the NLE ( noncoherent linear equalization ) schemes considered in [4, 9], the NLE scheme proposed here minimizes the variance of intersymol interference ( ISI ) in the equalizer output signal and thus, will be referred to as noncoherent minimum ISI equalization ( NMIE). The essential difference to all previously proposed noncoherent equalization schemes [3, 4, 5, 6, 9] is that a closed form solution for the equalizer coefficients exists. If an infinite length filter is employed, a zero forcing ( ZF ) equalizer results. For adaptation of the equalizer coefficients a modified least mean square ( LMS ) algorithm is presented. Simulations confirm the high performance of the proposed scheme and its robustness against frequency offset. 2. Transmission Model and Receiver Structure shows a block diagram of the discrete time transmission model. All signals are represented by their plex baseband equivalents. For simplicity, only T spaced equalizers are considered here. The transmitted MAPSK symbols s[k] are given by s[k] s[k] R[k]b[k], k? , with absolute amplitude symbol R[k], R[k]?{R1,? ,Rz} (z=1,., R[k]≡ 1, k? , for MDPSK) and absolute phase symbol b[k]? { ? ?2 / /j M Ze ?? |? ?{0,1,? ,M/Z1}}. For convenience, 2{ [ ] }sk? ({}?? denotes expectation) is normalized to unity. s[k] is obtained from the MDAPSK symbol . ? s[k] ? R[k]a[k] (1) via differential encoding. s[k] = ? s[k] s[k1] (2) ? R[k]=R[k]/R[k1] ( ? R[k]≡ 1, k? , for MDPSK ) and a[k]= b[k]/b[k1], denote the amplitude and the phase difference symbol, respectively. The most popular MDAPSK signaling format is 16DAPSK (16 star QAM) and will be used for the simulations presented in Section 5. Here, Z=2 is valid, . , there are two different amplitude levels R1 and R2,R1＜ R2, and the amplitude difference symbol, while three information bits are Gray mapped to the phase difference symbol. ? R[k]=1 and ? R[k]=? R ( if R[k1] = R1 ) or ? R[k]=1/? R ( if R[k1] = R2 ) is valid if the corresponding information bit is equal to zero and one, respectively. The MAPSK symbols s[k] are transmitted over an ISI producing channel with unknown, constant phase shift θ . The discrete time receiver input filter, can be expressed as 陜西理工學(xué)院畢業(yè)設(shè)計 第 17 頁 共 45頁 01[ ] [ ] [ ]hi Lr k e h s k n k??? ???? ? ?? , (3) where h? , 0? ? ? Lh1 are the coefficients of the bined discrete time impulse response of the cascade of transmit filter, channel, and receiver input filter, its length is denoted by Lh. For the receiver input filter, we assume a square root Nyquist frequency response. Thus, the zero mean plex Gaussian noise []n? is white. Due to an appropriate normalization, the noise variance is 22 0{ [ ] } / sn n k N E????. Es and N0 are the mean received energy per symbol and the single sided power spectral density of the underlying passband noise process, respectively. The equalizer output symbol q[k] may be written as 01[ ] [ ]cLq k c r k???????? 0000011[ ] [ ] [ ]chiivvhLLe g k s k k e g s k c n k????????????? ? ? ? ? ??? (4) where c? are the equalizer coefficients and 01vLcg c h? ? ?????? ? (5) are the coefficients of the bined impulse response of overall channel and equalizer。 Lc is the equalizer length. The decision delay k0 should be optimized since it can affect performance significantly. The next stage of the proposed receiver is a DF DD [7, 8]。 which determines an estimate ? ?s [k k0] for the transmitted symbol ? s[k k0] based on a reference symbol 1101 111201 11[]? [][ 1 ]1? []Nr e f Nqks k kqks k k?? ??? ???????? ???? ??? ? ???? ? ?? ?? ?。 (6) where N, 2N? is the number of equalizer output symbols used for determination of ? ?s [k k0] (cf. (7) ). The decision variable for estimation of ? ?s [k k0] is given by [][ ] .[ 1]refqkdk qk? ? (7)