DPCM编码MATLAB实现.docx

1、DPCM编码MATLAB实现DPCM编码MATLAB实现 %本文是数字图像处理的一个源程序 %实现的功能是DPCM编码 %DPCM编码,简称差值编码,是对模拟信号幅度抽样的差值进行量化编码的调制方式 %本程序实现一阶/二阶/三阶/四阶DPCM数字信号预测 %一阶/二阶/三阶/四阶预测的区别不仅在于信号的清晰度,而更重要在于 %阶数越高,图像越光滑. clc clear close all; %从D盘导入图片,以学校风光图片为例实现DPCM I03=imread(d:shuxuejianmo.bmp); %把RGB图像转化为灰度图像 I02=rgb2gray(I03); I=double(I02

2、); fid1=fopen(mydata1.dat,w); fid2=fopen(mydata2.dat,w); fid3=fopen(mydata3.dat,w); fid4=fopen(mydata4.dat,w); m,n=size(I); %对预测信号将边缘锁定,防止程序运行时抓不到数据 J1=ones(m,n); J1(1:m,1)=I(1:m,1); J1(1,1:n)=I(1,1:n); J1(1:m,n)=I(1:m,n); J1(m,1:n)=I(m,1:n); J2=ones(m,n); J2(1:m,1)=I(1:m,1); J2(1,1:n)=I(1,1:n); J2(

3、1:m,n)=I(1:m,n); J2(m,1:n)=I(m,1:n); J3=ones(m,n); J3(1:m,1)=I(1:m,1); J3(1,1:n)=I(1,1:n); J3(1:m,n)=I(1:m,n); J3(m,1:n)=I(m,1:n); J4=ones(m,n); J4(1:m,1)=I(1:m,1); J4(1,1:n)=I(1,1:n); J4(1:m,n)=I(1:m,n); J4(m,1:n)=I(m,1:n); %一阶DPCM编码 for k=2:m-1 for l=2:n-1 J1(k,l)=I(k,l)-I(k,l-1); end end J1=round

4、(J1); cont1=fwrite(fid1,J1,int8); cc1=fclose(fid1); %二阶DPCM编码 for k=2:m-1 for l=2:n-1 J2(k,l)=I(k,l)-(I(k,l-1)/2+I(k-1,l)/2); end end J2=round(J2); cont2=fwrite(fid2,J2,int8); cc2=fclose(fid2); %三阶DPCM编码 for k=2:m-1 for l=2:n-1 J3(k,l)=I(k,l)-(I(k,l-1)*(4/7)+I(k-1,l)*(2/7)+I(k-1,l-1)*(1/7); end end

5、J3=round(J3); cont3=fwrite(fid3,J3,int8); cc3=fclose(fid3); %四阶DPCM编码 for k=2:m-1 for l=2:n-1 J4(k,l)=I(k,l)-(I(k,l-1)/2+I(k-1,l)/4+I(k-1,l-1)/8+I(k-1,l+1)/8); end end J4=round(J4); cont4=fwrite(fid4,J4,int8); cc4=fclose(fid4); %= %以上是DPCM编码的编码过程,为了使程序具有连贯性,将编码和解码放在同一个M文件目录下 %= %以下是DPCM解码 fid1=fopen

6、(mydata1.dat,r); fid2=fopen(mydata2.dat,r); fid3=fopen(mydata3.dat,r); fid4=fopen(mydata4.dat,r); I11=fread(fid1,cont1,int8); I12=fread(fid2,cont2,int8); I13=fread(fid3,cont3,int8); I14=fread(fid4,cont4,int8); tt=1; for l=1:n for k=1:m I1(k,l)=I11(tt); tt=tt+1; end end tt=1; for l=1:n for k=1:m I2(k

7、,l)=I12(tt); tt=tt+1; end end tt=1; for l=1:n for k=1:m I3(k,l)=I13(tt); tt=tt+1; end end tt=1; for l=1:n for k=1:m I4(k,l)=I14(tt); tt=tt+1; end end I1=double(I1); I2=double(I2); I3=double(I3); I4=double(I4); J1=ones(m,n); J1(1:m,1)=I1(1:m,1); J1(1,1:n)=I1(1,1:n); J1(1:m,n)=I1(1:m,n); J1(m,1:n)=I1(

8、m,1:n); J2=ones(m,n); J2(1:m,1)=I2(1:m,1); J2(1,1:n)=I2(1,1:n); J2(1:m,n)=I2(1:m,n); J2(m,1:n)=I2(m,1:n); J3=ones(m,n); J3(1:m,1)=I3(1:m,1); J3(1,1:n)=I3(1,1:n); J3(1:m,n)=I3(1:m,n); J3(m,1:n)=I3(m,1:n); J4=ones(m,n); J4(1:m,1)=I4(1:m,1); J4(1,1:n)=I4(1,1:n); J4(1:m,n)=I4(1:m,n); J4(m,1:n)=I4(m,1:n)

9、; %一阶解码 for k=2:m-1 for l=2:n-1 J1(k,l)=I1(k,l)+J1(k,l-1); end end cc1=fclose(fid1); J1=uint8(J1); %二阶解码 for k=2:m-1 for l=2:n-1 J2(k,l)=I2(k,l)+(J2(k,l-1)/2+J2(k-1,l)/2); end end cc2=fclose(fid2); J2=uint8(J2); %三阶解码 for k=2:m-1 for l=2:n-1 J3(k,l)=I3(k,l)+(J3(k,l-1)*(4/7)+J3(k-1,l)*(2/7)+J3(k-1,l-

10、1)*(1/7); end end cc3=fclose(fid3); J3=uint8(J3); %四阶解码 for k=2:m-1 for l=2:n-1 J4(k,l)=I4(k,l)+(J4(k,l-1)/2+J4(k-1,l)/4+J4(k-1,l-1)/8+J4(k-1,l+1)/8); end end cc4=fclose(fid4); J4=uint8(J4); %分区画图 figure(1) subplot(3,2,1); imshow(I03); %隐藏坐标轴和边框,以免坐标轴与标题重叠 axis off box off title(原始图像,fontsize,11,fon

11、tname,隶体); subplot(3,2,2); imshow(I02); axis off box off title(灰度图像,fontsize,11,fontname,隶体); subplot(3,2,3); imshow(J1); axis off box off title(一阶预测,fontsize,11,fontname,隶体); subplot(3,2,4); imshow(J2); axis off box off title(二阶预测,fontsize,11,fontname,隶体); subplot(3,2,5); imshow(J3); axis off box o

12、ff title(三阶预测,fontsize,11,fontname,隶体); subplot(3,2,6); imshow(J4); axis off box off title(四阶预测,fontsize,11,fontname,隶体); % 学会DPCM编码必须彻底了解DPCM编码原理 % DPCM编码是数字图形处理的一项应用 % Removing all variables, functions, and MEX-files from memory, leaving the % workspace empty. clear all % Deleting all figures whos

13、e handles are not hidden. close all % Deleting all figures including those with hidden handles. close all hidden % Clearing all input and output from the Command Window display giving us a clean screen. clc % Opening the file TEOTH.mp3 in the read access mode. fid = fopen (TEOTH.mp3,r); % Generating

14、 the input signal m(t) by reading the binary data in 16 bit % integer format from the specified file and writing it into a matrix % m(t) and the number of elements successfully read is returned into an % output argument count. m,count = fread (fid,int16); % Redefining the count for efficiency. count

15、 = 8500; % Setting the sampling frequency. % because the audio signal has a maximum frequency of 4K and according to % Nyquist criteria, we get the following sampling frequency. Fs = 8000; % Setting the sampling instant. Ts = 1; % Setting the number of samples to be used. No_Samples = (2*Fs)+Ts; % D

16、efine the time vector for the calculations. time = 1:Fs/64; % Calculating maximum value of the input signal m(t). Mp = max (m) % Setting number of bits in a symbol. bits = 5; % Number of levels of uniform quantization. levels = 2bits; % Calculating the bit rate. bit_rate = 8000*bits; % Since the DPC

17、M is implemented by the linear predictor (transversal % predictor) Hence setting up the prediction coefficient alpha. alpha = 0.45; % Transmitting the difference. % Since there is no estimated value before the first sample so we get diff_sig(1) = m(1); % Calculating the rest of the values of the dif

18、ference signal with the help % of coefficient. for k = 2:count, diff_sig(k) = m(k) - alpha*m(k-1); end % Calculating maximum value of the input signal diff_sig(t),i.e, to be % quantized. Dp = max (diff_sig) % Calculating the step size of the quantization. step_size = (2*Mp)/levels % Quantizing the d

19、ifference signal. for k = 1:No_Samples, samp_in(k) = m(k*Ts); quant_in(k) = samp_in(k)/step_size; error(k) = (samp_in(k) - quant_in(k)/No_Samples; end % quant_in = diff_sig/step_size; % Indicating the sign of the input signal m(t) and calculating the % quantized signal quant_out. signS = sign (m); q

20、uant_out = quant_in; for i = 1:count, S(i) = abs (quant_in(i) + 0.5; quant_out(i) = signS(i)*round(S(i)*step_size; end % Decoding the signal using the quantized difference signal. s_out = quant_out; s_out(1) = quant_out(1); for k = 2:count, s_out(k) = quant_out(k) + alpha*s_out(k-1); end % Calculati

21、ng the quantization noise Nq. Nq = (step_size)2)/12)*(Mp/Dp)2) % Calculating signal to noise ratio SNR. SNR = 1.5*(levels2) SNR_db = 10*log10(SNR) % Plotting the input signal m(t). %figure; subplot(4,1,1); plot(time,m(time),time,s_out(time),r); title(Input Speech Signal); xlabel(Time); ylabel(m(t);

22、grid on; % Plotting the quantized signal quant_in(t). %figure; subplot(4,1,2); stem(time,quant_in(time),r); title(Quantized Speech Signal); xlabel(Time); ylabel(Levels); grid on; % Plotting the DPCM signal s_out(t). %figure; subplot(4,1,3); plot(time,s_out(time); title(Decoded DPCM Speech Signal); x

23、label(Time); ylabel(Dq(t); grid on; % Plotting the error signal error(t). subplot(4,1,4); plot(time,error(time); title(Error Signal); xlabel(Time); ylabel(Error(t); grid on; % Removing all variables, functions, and MEX-files from memory, leaving the % workspace empty. clear all clear all close all %

24、Solicitatipo de se馻l a muestrear opcion = input(Escriba 1 siquierecodificar se馻l de audio o 2 siquierecodificarotra: ) %abre se馻l seleccionadapor el usuario if opcion=1; t_input = input(escriba el tiempo en segundosquedeseegrabar el audio: ); m = wavrecord(t_input*30000,30000,int16); %fid = fopen (s

25、onido.wav,r); %m = fread (fid,int16); ini_cuenta = 10; end if opcion=2; t = input(escriba el vector de tiempopara la se馻l: ); m = input(escriba la se馻l quedeseecodificar f(t): ); ini_cuenta = 2; end %Solicitafrecuencia de muestreo y niveles de cuantizacion Fs = input(escriba la frecuencia de muestre

26、o: ); levels = input(escriba los niveles de cuantizacion: ); Mp = max (m) %Calcula el nivel m醲imo de la se馻l step_size = (Mp*2)/levels %Incremento entre cadanivel de cuant particion = -Mp:step_size:Mp; %vector de particion (cuant) %particion = 0:step_size:2*Mp; Ts = 1; longitud_m = length(m); inc_mu

27、estreo = longitud_m/Fs; red_inc_muestreo = floor(inc_muestreo); No_samples = (red_inc_muestreo*Fs)+1; %Numero de muestras, %Muestreo for k=ini_cuenta:No_samples if k = ini_cuenta samp_in(k-1) = 0; ind_pcm = 1 end residuo = rem(k,red_inc_muestreo); if residuo = 0 samp_in(k) = m(k); elseifresiduo = 0

28、samp_in(k) = samp_in(k-1); end end %Cuantizacion quant = quantiz(samp_in,particion); pcm_cad = dec2bin(quant) ind_pcm = 1; %Genera codigobinario de PCM for h=ini_cuenta:No_samples residuo = rem(h,red_inc_muestreo); if residuo = 0 PCM(ind_pcm) = str2num(pcm_cad(h,:); ind_pcm = ind_pcm+1; end end subp

29、lot(2,2,1); plot(m); title(se馻l anal骻ica); xlabel(tiempo); ylabel(amplitud); subplot(2,2,2); stairs(samp_in); title(se馻l muestreada); xlabel(tiempo); ylabel(amplitud); subplot(2,2,3); plot(quant); title(se馻l cuantizada); xlabel(tiempo); ylabel(niveles de cuantizacion); DPCM预测编码的MATLAB原代码: 收藏 DPCM预测编码原代码: i1=imread(3.jpg); i1=rgb2gray(i1); i1=imcrop(i1,20 20 160 160); i=double(i1); m,n=size(i); p=zeros(m,n); y=zeros(m,n); y(1:m,1)=i(1:m,1); p(1:m,1)=i(1:m,1);