-
Notifications
You must be signed in to change notification settings - Fork 0
/
inter_still_image.m
462 lines (444 loc) · 15.3 KB
/
inter_still_image.m
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
%lena_small = double(imread('lena_small.tif'));
Lena = double(imread('lena.tif'));
%% load image
for i = 20:40
name = ['foreman00',num2str(i),'.bmp'];
image(:,:,:,i-19) = double(imread(name));
rgbimage(:,:,:,i-19) = image(:,:,:,i-19);
image(:,:,:,i-19) = ictRGB2YCbCr(image(:,:,:,i-19));
end
%% initial parameters
[len,col,dem,fr] = size(image);
len_mv = [len/8*col/8];
%form = double(imread('foreman0020.bmp'));
EOB = 4000;
%scales = 1:0.6:1; % quantization scale factor, for E(4-1), we just evaluate scale factor of 1
scales = [0.07, 0.2, 0.4, 0.8, 1.0, 1.5, 2, 3, 4, 4.5];
bitPerPixel = zeros(numel(scales),1);
PSNR_mean = zeros(numel(scales),1);
max_min = -1000;
%% initial coefficients
for scaleIdx = 1 : numel(scales)
min_k = zeros(1,21);
len_k = zeros(1,21);
qScale = scales(scaleIdx);
bytestream = cell(21);
bytestream_mv = cell(20);
PSNR = zeros(1,21);
% k_small = IntraEncode(lena_small, qScale);
%% encoding decoding first image
fram_1 = image(:,:,:,1);
[k_1,im_zigz1] = IntraEncode(fram_1,qScale,EOB);
min_k(1) = min(k_1);
len_k(1) = length(k_1);
im_pmf = stats_marg(k_1,min(k_1):max(k_1));
[Binarytree_first,HuffCode_first,BinCode_first,CodeLengths_first] = buildHuffman(im_pmf);
bytestream{1} = enc_huffman_new(k_1-min(k_1)+1,BinCode_first,CodeLengths_first);
k_rec_first = dec_huffman_new(bytestream{1},Binarytree_first,len_k(1));
k_rec_first = k_rec_first+min_k(1)-1;
im_rec(:,:,:,1) = IntraDecode(k_rec_first,size(fram_1),qScale,EOB);
im_rec_rgb(:,:,:,1) = ictYCbCr2RGB(im_rec(:,:,:,1));
min_mv = zeros(20);
PSNR(1) = calcPSNR(rgbimage(:,:,:,1),im_rec_rgb(:,:,:,1));
bits = numel(bytestream{1})*8;
%bit_rate = bits/(len*col*dem/3)
bitrate_im = zeros(21,1);
PSNR_im = zeros(21,1);
bitrate_im(1) = bits/(len*col);
PSNR_im(1) = PSNR(1);
%% encoding and decoding follow images.
for seq = 2:21
fram = image(:,:,:,seq);
ref_fram = im_rec(:,:,:,seq-1);
fram_1dim = fram(:,:,1);
ref_fram_1dim = ref_fram(:,:,1);
mv = SSD(ref_fram_1dim,fram_1dim);
res = fram-SSD_rec(ref_fram,mv); %the first res
[res_k,im_zigz] = IntraEncode(res, qScale,EOB);
min_k(seq) = min(res_k);
len_k(seq) = length(res_k);
min_mv(seq-1) = min(mv(:));
if seq == 2 % use the second image to construct huffman code.
mv_pmf = stats_marg(mv,1:81);
[Binarytree_mv,HuffCode_mv,BinCode_mv,CodeLengths_mv] = buildHuffman(mv_pmf);
res_pmf = stats_marg(res_k,max_min:4000);
[Binarytree_res,HuffCode_res,BinCode_res,CodeLengths_res] = buildHuffman(res_pmf);
end
%seq
bytestream{seq} = enc_huffman_new(res_k-max_min+1,BinCode_res,CodeLengths_res); % huffman encode res.
bytestream_mv{seq-1} = enc_huffman_new(mv(:)-min_mv(seq-1)+1,BinCode_mv,CodeLengths_mv); %huffman encode the vector
k_rec = dec_huffman_new(bytestream{seq},Binarytree_res,len_k(seq));
k_rec = k_rec+max_min-1;
res_rec = IntraDecode(k_rec, size(fram_1),qScale,EOB);
mv_rec = dec_huffman_new(bytestream_mv{seq-1},Binarytree_mv,len_mv);
mv_rec = mv_rec+min_mv(seq-1)-1;
mv_rec = reshape(mv_rec,[len/8,col/8]);
im_rec_current = SSD_rec(ref_fram,mv_rec)+res_rec;
im_rec(:,:,:,seq) = im_rec_current;
im_rec_rgb(:,:,:,seq) = ictYCbCr2RGB(im_rec_current);
PSNR(seq) = calcPSNR(rgbimage(:,:,:,seq),im_rec_rgb(:,:,:,seq));
bits = bits+numel(bytestream{seq})*8+numel(bytestream_mv{seq-1})*8;
%bit_rate = (numel(bytestream{seq})*8+numel(bytestream_mv{seq-1})*8)/(len*col)
[bitrate_im(seq),PSNR_im(seq)] = im_code(qScale,fram,EOB);
seq
end
PSNR_image(scaleIdx) = sum(PSNR_im)/21;
bitrate_image(scaleIdx) = sum(bitrate_im)/21;
PSNR_mean(scaleIdx) = mean(PSNR(:));
bitPerPixel(scaleIdx) = bits/(numel(image)/3);
fprintf('QP: %.1f bit-rate: %.2f bits/pixel PSNR: %.2fdB\n', qScale, bitPerPixel(scaleIdx), PSNR_mean(scaleIdx))
end
plot(bitPerPixel,PSNR_mean,'x-');
hold on
title('dr plot');
xlabel('bitPerPixel');
ylabel('PSNR');
plot(bitrate_image,PSNR_image,'x-');
legend('inter prediction','still image code')
%% put all used sub-functions here.
function [Bitrate,PSNR_IM] = im_code(qScale,Lena,EOB)
[k,im_zigz] = IntraEncode(Lena, qScale,EOB);
im_pmf = stats_marg(k,max(k)-min(k)+1);
[Binarytree,HuffCode,BinCode,CodeLengths] = buildHuffman(im_pmf);
bytestream = enc_huffman_new(k-min(k)+1,BinCode,CodeLengths);
%% use pmf of k_small to build and train huffman table
%your code here
%% use trained table to encode k to get the bytestream
% your code here
k_rec = dec_huffman_new(bytestream,Binarytree,length(k));
k_rec = k_rec+min(k)-1;
Bitrate = (numel(bytestream)*8) / (numel(Lena)/3);
%% image reconstruction
I_rec = IntraDecode(k_rec, size(Lena),qScale,EOB);
I_rec = ictYCbCr2RGB(I_rec);
Lena = ictYCbCr2RGB(Lena);
PSNR_IM = calcPSNR(I_rec,Lena);
%PSNR(scaleIdx) = calcPSNR(Lena, I_rec);
%fprintf('QP: %.1f bit-rate: %.2f bits/pixel PSNR: %.2fdB\n', qScale, bitPerPixel(scaleIdx), PSNR(scaleIdx))
end
function dst = IntraDecode(image, img_size , qScale,EOB)
% Function Name : IntraDecode.m
% Input : image (zero-run encoded image, 1xN)
% img_size (original image size)
% qScale(quantization scale)
% Output : dst (decoded image)
image_rund = ZeroRunDec_EoB(image,EOB);
image_rund = reshape(image_rund, [img_size(1)/8*64,img_size(2)/8*img_size(3)]);
image_dezigz = blockproc(image_rund, [64,3],@(block_struct) DeZigZag8x8(block_struct.data));
image_dequant = blockproc(image_dezigz,[8,8],@(block_struct) DeQuant8x8(block_struct.data,qScale));
image_idc = blockproc(image_dequant,[8,8],@(block_struct) IDCT8x8(block_struct.data));
%dst = ictYCbCr2RGB(image_idc);
dst = image_idc;
end
function [dst,im_zigz] = IntraEncode(image, qScale,EOB)
% Function Name : IntraEncode.m
% Input : image (Original RGB Image)
% qScale(quantization scale)
% Output : dst (sequences after zero-run encoding, 1xN)
%im_tran = ictRGB2YCbCr(image);
im_tran = image;
[row,col,dem] = size(im_tran);
im_dct = blockproc(im_tran,[8,8],@(block_struct) DCT8x8(block_struct.data));
im_quant = blockproc(im_dct,[8,8],@(block_struct) Quant8x8(block_struct.data,qScale));
im_zigz = blockproc(im_quant,[8,8],@(block_struct) ZigZag8x8(block_struct.data));
im_zigz = im_zigz(:);
dst = ZeroRunEnc_EoB(im_zigz,EOB);
end
%% and many more functions
function coeff = DCT8x8(block)
% Input : block (Original Image block, 8x8x3)
%
% Output : coeff (DCT coefficients after transformation, 8x8x3)
dct8 = zeros(8,8);
[row,col,dem] = size(block);
coeff = zeros(row,col,dem);
for ii = 1:8
for jj = 1:8
k = ii-1;
n = jj-1;
if k == 0
dct_t = 1/sqrt(8);
else
dct_t = sqrt(2/8)*cos(pi*(2*n+1)*k/(2*8));
end
dct8(ii,jj) = dct_t;
end
end
for ii = 1:dem
coeff(:,:,ii) = dct8*block(:,:,ii)*dct8.';
end
end
function block = IDCT8x8(coeff)
% Function Name : IDCT8x8.m
% Input : coeff (DCT Coefficients) 8*8*3
% Output : block (original image block) 8*8*3
[row,col,ded] = size(coeff);
dct8 = zeros(8,8);
for i = 1:8
for j = 1:8
k = i-1;
n = j-1;
if k == 0
dctte = 1/sqrt(8);
else
dctte = sqrt(2/8)*cos((2*n+1)*pi*k/(2*8));
end
dct8(i,j) = dctte;
end
end
for i = 1 :ded
block(:,:,i) = dct8'*coeff(:,:,i)*dct8;
end
end
function zze = ZeroRunEnc_EoB(zz,EOB)
%EOB = 1000;
% Input : zz (Zig-zag scanned sequence, 1xN)
% EOB (End Of Block symbol, scalar)
%
% Output : zze (zero-run-level encoded sequence, 1xM)
len = length(zz);
zze = [];
for rec = 1:len/64
if rec ~= len/64
zz_t = zz((rec-1)*64+1:rec*64);
else
zz_t = zz((rec-1)*64+1:len);
end
[lens] = length(zz_t);
k = 1;
zze_t = [];
flag = 0;
m = 0;
while k <= lens
if zz_t(k) == 0
count = -1;
m = k;
while (zz_t(m) == 0 )
if m == 1
flag1 = 1;
end
count = count+1;
m = m+1;
if m > lens
flag = 1;
break
end
end
zze_t = [zze_t,0,count];
k = m-1;
else
zze_t = [zze_t,zz_t(k)];
end
k = k+1;
end
len2 = length(zze_t);
if flag == 1
zze_t = zze_t(1:len2-1);
zze_t(end) = EOB;
end
zze = [zze,zze_t];
end
end
function dst = ZeroRunDec_EoB(src,EOB)
%EoB = 1000;
% Function Name : ZeroRunDec1.m zero run level decoder
% Input : src (zero run encoded sequence 1xM with EoB signs)
% EoB (end of block sign)
%
% Output : dst (reconstructed zig-zag scanned sequence 1xN)
k = 0;
len_src = length(src);
m = 1;
dst = [];
while k < len_src
k = k+1;
if m == 65
m = 1;
end
if src(k) == EOB
dst(end+1:end+64-m+1) = 0;
m = 1;
continue
end
if src(k) == 0
dst(end+1:end+src(k+1)+1) = 0;
m = m+src(k+1)+1;
k = k+1;
continue
end
dst(end+1) = src(k);
m = m+1;
end
size(dst);
size(src);
end
function pmf = stats_marg(image,range)
[cc] = hist(image(:),range);
pmf = cc/sum(cc);
end
function yuv = ictRGB2YCbCr(rgb)
% Input : rgb (Original RGB Image)
% Output : yuv (YCbCr image after transformation)
% YOUR CODE HERE
yuv(:,:,1) = 0.299.*rgb(:,:,1)+0.587.*rgb(:,:,2)+0.114.*rgb(:,:,3);
yuv(:,:,2) = -0.169.*rgb(:,:,1)-0.331.*rgb(:,:,2)+0.5.*rgb(:,:,3);
yuv(:,:,3) = 0.5.*rgb(:,:,1)-0.419.*rgb(:,:,2)-0.081.*rgb(:,:,3);
end
function rgb = ictYCbCr2RGB(yuv)
% Input : yuv (Original YCbCr image)
% Output : rgb (RGB Image after transformation)
% YOUR CODE HERE
Y = yuv(:,:,1);
Cb = yuv(:,:,2);
Cr = yuv(:,:,3);
r = Y+1.402.*Cr;
g = Y-0.344.*Cb-0.714.*Cr;
b = Y+1.772.*Cb;
rgb(:,:,1) = r;
rgb(:,:,2) = g;
rgb(:,:,3) = b;
end
function PSNR = calcPSNR(Image, recImage)
MSE = calcMSE(Image,recImage);
PSNR = 10*log10((2^8-1)^2/MSE);
end
function MSE = calcMSE(Image, recImage)
[len,col,de] = size(Image);
diff = Image - recImage;
diff = diff.^2;
sumd = sum(diff(:));
MSE = sumd/(len*col*de);
end
function quant = Quant8x8(dct_block, qScale)
% Input : dct_block (Original Coefficients, 8x8x3)
% qScale (Quantization Parameter, scalar)
%
% Output : quant (Quantized Coefficients, 8x8x3)
L = [16,11,10,16,24,40,51,61;12,12,14,19,26,58,60,55;14,13,16,24,40,57,69,56;14,17,22,29,...
51,87,80,62;18,55,37,56,68,109,103,77;24,35,55,64,81,104,113,92;49,64,78,87,103,121,120,101;...
72,92,95,98,112,100,103,99];
c = [17,18,24,47,99,99,99,99;18,21,26,66,99,99,99,99;24,13,56,99,99,99,99,99;47,66,99,99,99,99,99,99;99,99,...
99,99,99,99,99,99;99,99,99,99,99,99,99,99;99,99,99,99,99,99,99,99;99,99,99,99,99,99,99,99];
quant = zeros(8,8,3);
L = qScale.*L;
c = qScale.*c;
quant(:,:,1) = round(dct_block(:,:,1)./L);
quant(:,:,2) = round(dct_block(:,:,2)./c);
quant(:,:,3) = round(dct_block(:,:,3)./c);
end
function dct_block = DeQuant8x8(quant_block, qScale)
% Function Name : DeQuant8x8.m
% Input : quant_block (Quantized Block, 8x8x3)
% qScale (Quantization Parameter, scalar)
%
% Output : dct_block (Dequantized DCT coefficients, 8x8x3)
L = [16,11,10,16,24,40,51,61;12,12,14,19,26,58,60,55;14,13,16,24,40,57,69,56;14,17,22,29,...
51,87,80,62;18,55,37,56,68,109,103,77;24,35,55,64,81,104,113,92;49,64,78,87,103,121,120,101;...
72,92,95,98,112,100,103,99];
c = [17,18,24,47,99,99,99,99;18,21,26,66,99,99,99,99;24,13,56,99,99,99,99,99;47,66,99,99,99,99,99,99;99,99,...
99,99,99,99,99,99;99,99,99,99,99,99,99,99;99,99,99,99,99,99,99,99;99,99,99,99,99,99,99,99];
L = qScale.*L;
c = qScale.*c;
dct_block(:,:,1) = quant_block(:,:,1).*L;
dct_block(:,:,2) = quant_block(:,:,2).*c;
dct_block(:,:,3) = quant_block(:,:,3).*c;
end
function zz = ZigZag8x8(quant)
% Input : quant (Quantized Coefficients, 8x8x3)
%
% Output : zz (zig-zag scaned Coefficients, 64x3)
zigz = [1 2 6 7 15 16 28 29;
3 5 8 14 17 27 30 43;
4 9 13 18 26 31 42 44;
10 12 19 25 32 41 45 54;
11 20 24 33 40 46 53 55;
21 23 34 39 47 52 56 61;
22 35 38 48 51 57 60 62;
36 37 49 50 58 59 63 64];
for k = 1:3
quantt = quant(:,:,k);
zz(zigz(:),k) = quantt(:);
end
end
function coeffs = DeZigZag8x8(zz)
% Function Name : DeZigZag8x8.m
% Input : zz (Coefficients in zig-zag order)
%
% Output : coeffs(DCT coefficients in original order)
zigz = [ 1 2 6 7 15 16 28 29;
3 5 8 14 17 27 30 43;
4 9 13 18 26 31 42 44;
10 12 19 25 32 41 45 54;
11 20 24 33 40 46 53 55;
21 23 34 39 47 52 56 61;
22 35 38 48 51 57 60 62;
36 37 49 50 58 59 63 64];
for k = 1:3
coeffs_t = zz(zigz(:),k);
coeffs_t = reshape(coeffs_t,[8,8]);
coeffs(:,:,k) = coeffs_t;
end
end
function motion_vectors_indices = SSD(ref_image, image)
% Input : ref_image(Reference Image, size: height x width)
% image (Current Image, size: height x width)
%
% Output : motion_vectors_indices (Motion Vector Indices, size: (height/8) x (width/8) x 1 )ref
range = 4;
ref_image = padarray(ref_image,[range,range],0,'both');
%ref_image = padarray(ref_image,[7,7],0,'post');
[len,col] = size(image);
tx = len/8;
ty = col/8;
for i = 1:tx
for j = 1:ty
block = image((i-1)*8+1:i*8,(j-1)*8+1:j*8);
search = ref_image((i-1)*8+1:(i-1)*8+1+range+range+7,(j-1)*8+1:(j-1)*8+1+range+range+7);
mv = searchmv(search,block);
mv_ind = (mv(1)-1+5)*9+mv(2)+5;
motion_vectors_indices(i,j) = mv_ind;
end
end
end
function mv = searchmv(search,block)
best_ssd = inf;
for i = 1:9
for j = 1:9
search_block = search(i:i+7,j:j+7);
ssd_current = sum((search_block-block).^2,'all');
if ssd_current<best_ssd
best_ssd = ssd_current;
mv = [i,j]-[5,5];
end
end
end
end
function rec_image = SSD_rec(ref_image, motion_vectors)
% Input : ref_image(Reference Image, YCbCr image)
% motion_vectors
%
% Output : rec_image (Reconstructed current image, YCbCr image)
[len,col,dem] = size(ref_image);
rec_image = zeros(len,col,3);
ref_image = padarray(ref_image,[4,4],'both');
tx = len/8;
ty = col/8;
for i = 1:dem
for j = 1:tx
for k = 1:ty
search = ref_image((j-1)*8+1:(j-1)*8+1+4+4+7,(k-1)*8+1:(k-1)*8+1+4+4+7,i);
mv_ind = motion_vectors(j,k);
mv = zeros(1,2);
if mod(mv_ind,9) == 0
mv(1) = mv_ind/9;
else
mv(1) = floor(mv_ind/9)+1;
end
mv(2) = mv_ind-(mv(1)-1)*9;
search_block = search(mv(1):mv(1)+7,mv(2):mv(2)+7);
block = search_block;
rec_image((j-1)*8+1:j*8,(k-1)*8+1:k*8,i) = block;
end
end
end
end