‫ﺑﺎﺳﻤﻪ ﺗﻌﺎﻟﯽ‬
‫ﺳﯿﺴﺘﻢﻫﺎي ﭼﻨﺪرﺳﺎﻧﻪاي )‪(40-342‬‬
‫داﻧﺸﮑﺪه ﻣﻬﻨﺪﺳﯽ ﮐﺎﻣﭙﯿﻮﺗﺮ‬
‫ﺗﺮم ﺑﻬﺎر ‪1387‬‬
‫دﮐﺘﺮ ﺣﻤﯿﺪرﺿﺎ رﺑﯿﻌﯽ‬
‫ﺗﮑﻠﯿﻒ ﺷﻤﺎره ‪ :5‬ﻓﺸﺮده ﺳﺎزي ﺗﺼﻮﯾﺮ‬
‫‪ -1‬ﻣﻘﺪﻣﻪ‬
‫ﻳﻚ ﺗﺼﻮﻳﺮ ﺩﻳﺠﻴﺘﺎﻝ ﺑﻪ ﺩﺳﺖ ﺁﻣﺪﻩ ﺑﺎ ﻧﻤﻮﻧﻪﺑﺮﺩﺍﺭﻱ ﻭ ﻛﻮﺍﻧﺘﻴﺰﺍﺳﻴﻮﻥ ﻳﻚ ﻋﻜﺲ ﻣﻌﻤﻮﻟﻲ‪ ،‬ﻓﻀﺎﻱ ﺫﺧﻴﺮﻩﺳﺎﺯﻱ ﺯﻳﺎﺩﻱ ﺭﺍ ﻧﻴﺎﺯ ﺩﺍﺭﺩ‪ .‬ﺑﻪ ﻋﻨﻮﺍﻥ‬
‫ﻣﺜﺎﻝ‪ ،‬ﻳﻚ ﺗﺼﻮﻳﺮ ﺭﻧﮕﻲ ‪ ٥١٢*٥١٢‬ﻛﻪ ﺭﻧﮓ ﻫﺮ ﭘﻴﻜﺴﻞ ﺭﺍ ﺑﺎ ‪ ٢٤‬ﺑﻴﺖ ﻧﺸﺎﻥ ﻣﻲﺩﻫﺪ‪ ٧٦٨ ،‬ﻛﻴﻠﻮﺑﺎﻳﺖ ﺍﺯ ﺣﺠﻢ ﻳﻚ ﺩﻳﺴﻚ ﺭﺍ ﺍﺷﻐﺎﻝ ﻣﻲ‬
‫ﻛﻨﺪ ﻭ ﻋﻜﺴﻲ ﺑﺎ ﺍﻧﺪﺍﺯﻩ ﺩﻭ ﺑﺮﺍﺑﺮ ﻋﻜﺲ ﻗﺒﻠﻲ ﺑﺮ ﺭﻭﻱ ﻳﻚ ﺩﻳﺴﻜﺖ ﻧﺮﻡ ﺟﺎ ﻧﻤﻲﺷﻮﺩ‪ .‬ﺑﺮﺍﻱ ﺍﻧﺘﻘﺎﻝ ﭼﻨﻴﻦ ﺗﺼﻮﻳﺮﻱ ﺑﺎ ﻳﻚ ﻣﻮﺩﻡ ‪kbps‬‬
‫‪ ٢٨/٨‬ﺑﻪ ﺣﺪﻭﺩ ‪ ٤‬ﺩﻗﻴﻘﻪ ﺯﻣﺎﻥ ﻧﻴﺎﺯ ﺍﺳﺖ‪ .‬ﻫﺪﻑ ﺍﺯ ﻓﺸﺮﺩﻩ ﺳﺎﺯﻱ ﺗﺼﻮﻳﺮ ﻛﺎﻫﺶ ﻣﻴﺰﺍﻥ ﺩﺍﺩﻩﻫﺎﻱ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺑﺮﺍﻱ ﻧﺸﺎﻥ ﺩﺍﺩﻥ ﺗﺼﺎﻭﻳﺮ ﺩﻳﺠﻴﺘﺎﻝ‬
‫ﻭ ﺑﻨﺎﺑﺮﺍﻳﻦ ﻛﺎﻫﺶ ﻫﺰﻳﻨﻪ ﺍﻧﺘﻘﺎﻝ ﻭ ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﻣﻲﺑﺎﺷﺪ‪ .‬ﻓﺸﺮﺩﻩﺳﺎﺯﻱ ﺗﺼﻮﻳﺮ ﻧﻘﺶ ﻛﻠﻴﺪﻱ ﺩﺭ ﺑﺴﻴﺎﺭﻱ ﺍﺯ ﻛﺎﺭﺑﺮﺩﻫﺎﻱ ﻣﻬﻢ ﺩﺍﺭﺍﺳﺖ‪ .‬ﺍﻳﻦ‬
‫ﻛﺎﺭﺑﺮﺩﻫﺎ ﺷﺎﻣﻞ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺗﺼﻮﻳﺮ‪ ،‬ﻣﺨﺎﺑﺮﻩ ﺗﺼﻮﻳﺮ‪ ،‬ﺳﻨﺠﺶ ﺍﺯ ﺭﺍﻩ ﺩﻭﺭ )ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺗﺼﺎﻭﻳﺮ ﻣﺎﻫﻮﺍﺭﻩ ﺍﻱ ﺑﺮﺍﻱ ﻛﺎﺭﺑﺮﺩﻫﺎﻱ ﺁﺏ ﻭ ﻫﻮﺍ ﻳﺎ‬
‫ﻣﻨﺎﺑﻊ ﺯﻳﺮﺯﻣﻴﻨﻲ(‪ ،‬ﺗﺼﻮﻳﺮﺑﺮﺩﺍﺭﻱ ﺍﺳﻨﺎﺩ‪ ،‬ﺗﺼﻮﻳﺮ ﺑﺮﺩﺍﺭﻱ ﭘﺰﺷﻜﻲ‪ ،‬ﺍﻧﺘﻘﺎﻝ ﺗﻮﺳﻂ ﺩﻭﺭﻧﮕﺎﺭ‪ ،‬ﻛﻨﺘﺮﻝ ﻭﺳﺎﻳﻞ ﻧﻘﻠﻴﻪ ﻧﻈﺎﻣﻲ ﻭ ﻓﻀﺎﻳﻲ ﺍﺯ ﺭﺍﻩ ﺩﻭﺭ ﻭ‬
‫ﻛﻨﺘﺮﻝ ﺯﺑﺎﻟﻪﻫﺎﻱ ﺧﻄﺮﻧﺎﻙ ﻛﺎﺭﺑﺮﺩ ﺩﺍﺭﺩ‪ .‬ﺑﻪ ﻃﻮﺭ ﺧﻼﺻﻪ‪ ،‬ﻛﺎﺭﺑﺮﺩﻫﺎﻱ ﺭﻭﺯﺍﻓﺰﻭﻥ ﭘﺮﺩﺍﺯﺵ ﺗﺼﻮﻳﺮ‪ ،‬ﺑﻪ ﺩﺳﺘﻜﺎﺭﻱ‪ ،‬ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﻭ ﺍﻧﺘﻘﺎﻝ‬
‫ﺗﺼﺎﻭﻳﺮ ﺩﻭﺩﻭﻳﻲ‪ ،‬ﺳﻴﺎﻩ ﺳﻔﻴﺪ ﻳﺎ ﺭﻧﮕﻲ ﺑﻪ ﺷﻜﻠﻲ ﻛﺎﺭﺍ ﻭ ﻣﺆﺛﺮ ﺑﺴﺘﮕﻲ ﺩﺍﺭﺩ‪.‬‬
‫ﻳﻜﻲ ﺍﺯ ﻣﻬﻤﺘﺮﻳﻦ ﭘﻴﺸﺮﻓﺖﻫﺎ ﺩﺭ ﻓﺸﺮﺩﻩﺳﺎﺯﻱ ﺗﺼﻮﻳﺮ‪ ،‬ﺗﺄﺳﻴﺲ ﺍﺳﺘﺎﻧﺪﺍﺭﺩ ‪ JPEG‬ﺑﺮﺍﻱ ﻓﺸﺮﺩﻩﺳﺎﺯﻱ ﺗﺼﺎﻭﻳﺮ ﺭﻧﮕﻲ ﺑﻮﺩ‪ .‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ‬
‫ﺭﻭﺵ ‪ ،JPEG‬ﻳﻚ ﺗﺼﻮﻳﺮ ﺭﻧﮕﻲ ‪ ٢٤‬ﭘﻴﻜﺴﻞ‪/‬ﺑﻴﺖ ﻣﻲﺗﻮﺍﻧﺪ ﺑﻪ ‪ ١‬ﺗﺎ ‪ ٢‬ﭘﻴﻜﺴﻞ‪/‬ﺑﻴﺖ ﺗﻘﻠﻴﻞ ﭘﻴﺪﺍ ﻛﻨﺪ‪ .‬ﭼﻨﻴﻦ ﻛﺎﻫﺸﻲ‪ ،‬ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﻭ ﺍﻧﺘﻘﺎﻝ‬
‫ﺗﺼﺎﻭﻳﺮ ﻣﺎﻫﻮﺍﺭﻩﺍﻱ ﺭﺍ ﺑﺎ ﻫﺰﻳﻨﻪﺍﻱ ﻣﻌﻘﻮﻝ ﻣﻤﻜﻦ ﻣﻲﺳﺎﺯﺩ‪ ،‬ﺑﺪﻭﻥ ﺍﻳﻨﻜﻪ ﺁﺭﺗﻴﻔﻜﺖﻫﺎﻱ ﺁﺷﻜﺎﺭ ﻗﺎﺑﻞ ﻣﺸﺎﻫﺪﻩ ﺍﻳﺠﺎﺩ ﻛﻨﺪ‪ .‬ﺍﻳﻦ ﻛﺎﺭ ﻫﻤﭽﻨﻴﻦ‬
‫ﺍﻣﻜﺎﻥ ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﻳﻚ ﻋﻜﺲ ﺭﻧﮕﻲ ﺭﺍ ﺍﺯ ﺭﻭﻱ ﺍﻳﻨﺘﺮﻧﺖ ﺩﺭ ﺯﻣﺎﻧﻲ ﺣﺪﻭﺩ ﻳﻚ ﻟﺤﻈﻪ ﻓﺮﺍﻫﻢ ﻣﻲ ﻛﻨﺪ ﻭ ﺑﻨﺎﺑﺮﺍﻳﻦ ﺍﻧﺘﺸﺎﺭ ﻭ ﺗﺒﻠﻴﻐﺎﺕ ﺑﺮ ﺭﻭﻱ‬
‫ﺷﺒﻜﻪ ﺟﻬﺎﻧﻲ ﺍﻳﻨﺘﺮﻧﺖ ﺑﻪ ﻳﻚ ﻭﺍﻗﻌﻴﺖ ﺗﺒﺪﻳﻞ ﻣﻲﺷﻮﺩ‪ .‬ﻗﺒﻞ ﺍﺯ ﺍﺳﺘﺎﻧﺪﺍﺭﺩ ‪ ،JPEG‬ﺍﺳﺘﺎﻧﺪﺍﺭﺩﻫﺎﻱ ‪ G3‬ﻭ ‪ G4‬ﺑﺮﺍﻱ ﻓﺸﺮﺩﻩ ﺳﺎﺯﻱ ﺍﺳﻨﺎﺩ‬
‫ﺩﻭﺭﻧﮕﺎﺭ‪ ،‬ﻛﺎﻫﺶ ﺯﻣﺎﻥ ﺍﻧﺘﻘﺎﻝ ﻳﻚ ﺻﻔﺤﻪ ﻣﺘﻨﻲ ﺍﺯ ﺣﺪﻭﺩ ‪ ٦‬ﺩﻗﻴﻘﻪ ﺑﻪ ﻳﻚ ﺩﻗﻴﻘﻪ ﺭﺍ ﻣﻤﻜﻦ ﺳﺎﺧﺘﻪ ﺑﻮﺩ‪.‬‬
‫ﺩﺭ ﺍﻳﻦ ﺁﺯﻣﺎﻳﺶ‪ ،‬ﺍﺻﻮﻝ ﻓﺸﺮﺩﻩ ﺳﺎﺯﻱ ﺗﺼﻮﻳﺮ ﺑﺮﺍﻱ ﺗﺼﺎﻭﻳﺮ ﺩﻭﺩﻭﻳﻲ‪ ،‬ﺳﻴﺎﻩ ﺳﻔﻴﺪ ﻭ ﺭﻧﮕﻲ ﻣﻌﺮﻓﻲ ﺧﻮﺍﻫﺪ ﺷﺪ‪ .‬ﻓﺸﺮﺩﻩ ﺳﺎﺯﻱ ﻭﻳﺪﻳﻮ ﺩﺭ‬
‫ﺁﺯﻣﺎﻳﺶ ﺑﻌﺪﻱ ﺑﺮﺭﺳﻲ ﺧﻮﺍﻫﺪ ﺷﺪ‪.‬‬
‫‪ -٢‬ﺗﺌﻮري ﻫﺎ و ﺗﮑﻨﯿﮏ ﻫﺎي ﻓﺸﺮدهﺳﺎزي ﺗﺼﻮﯾﺮ‬
‫ﺑﻪ ﻃﻮﺭ ﻛﻠﻲ‪ ،‬ﺭﻭﺵ ﻫﺎﻱ ﻛﺪﻳﻨﮓ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺑﻪ ﺩﻭ ﺩﺳﺘﻪ ﺑﺪﻭﻥ ﺗﻠﻔﺎﺕ ﻭ ﺑﺎ ﺗﻠﻔﺎﺕ ﺗﻘﺴﻴﻢ ﺑﻨﺪﻱ ﺷﻮﻧﺪ‪ .‬ﺩﺭ ﻛﺪﻳﻨﮓ ﺑﺪﻭﻥ ﺗﻠﻔﺎﺕ‪ ،‬ﻣﻘﺎﺩﻳﺮ‬
‫ﻧﻤﻮﻧﻪ ﻫﺎﻱ ﺍﺻﻠﻲ ﺩﻭﺑﺎﺭﻩ ﺩﻗﻴﻘﺎﹰ ﺑﻪ ﺩﺳﺖ ﻣﻲ ﺁﻳﻨﺪ ﻭ ﻓﺸﺮﺩﻩﺳﺎﺯﻱ ﺑﻮﺳﻴﻠﻪ ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﺯﻭﺍﻳﺪ ﺁﻣﺎﺭﻱ ﺳﻴﮕﻨﺎﻝ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﺩ‪ .‬ﺩﺭ ﻛﺪﻳﻨﮓ ﺑﺎ‬
‫ﺗﻠﻔﺎﺕ‪ ،‬ﺳﻴﮕﻨﺎﻝ ﺍﺻﻠﻲ ﺗﺎ ﺣﺪﻭﺩﻱ ﺗﻐﻴﻴﺮ ﻣﻲ ﻛﻨﺪ ﻭﻟﻲ ﺑﻪ ﻧﺮﺥ ﻓﺸﺮﺩﻩ ﺳﺎﺯﻱ ﺑﺎﻻﺗﺮﻱ ﺩﺳﺖ ﻣﻲ ﻳﺎﺑﻴﻢ‪.‬‬
‫ﭘﺮﺩﺍﺯﺵ ﺑﻪ ﻛﺎﺭ ﮔﺮﻓﺘﻪ ﺷﺪﻩ ﺩﺭ ﺳﻴﺴﺘﻢ ﻛﺪﻳﻨﮓ ﺑﺎ ﺗﻠﻔﺎﺕ ﺩﺭ ﺷﻜﻞ ‪ ١‬ﻧﺸﺎﻥ ﺩﺍﺩﻩ ﺷﺪﻩ ﺍﺳﺖ‪ .‬ﻣﻲﺑﻴﻨﻴﻢ ﻛﻪ ﺳﻴﮕﻨﺎﻝ ﺩﺭ ﺍﺑﺘﺪﺍ ﺑﻪ ﻣﻴﺪﺍﻥ ﺗﺒﺪﻳﻞ‬
‫ﻣﻲ ﺭﻭﺩ ﻛﻪ ﺑﺮﺍﻱ ﻓﺸﺮﺩﻩ ﺳﺎﺯﻱ ﻣﻨﺎﺳﺐ ﺗﺮ ﺍﺳﺖ‪ .‬ﺳﭙﺲ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﺗﺒﺪﻳﻞ ﻛﻮﺍﻧﺘﻴﺰﻩ ﻣﻲﺷﻮﻧﺪ‪ .‬ﺩﺭ ﺍﻧﺘﻬﺎ‪ ،‬ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻛﻮﺍﻧﺘﻴﺰﻩ ﺷﺪﻩ ﺑﻪ ﺻﻮﺭﺕ‬
‫ﺑﺪﻭﻥ ﺗﻠﻔﺎﺕ ﺑﻪ ﺑﻴﺖ ﻫﺎﻱ ﺩﻭﺩﻭﻳﻲ ﻛﺪ ﻣﻲﺷﻮﻧﺪ‪ .‬ﺩﺭ ﻛﺪﻳﻨﮓ ﺑﺪﻭﻥ ﺗﻠﻔﺎﺕ‪ ،‬ﻣﺮﺣﻠﻪ ﻛﻮﺍﻧﺘﻴﺰﺍﺳﻴﻮﻥ ﺍﻧﺠﺎﻡ ﻧﻤﻲﺷﻮﺩ‪.‬‬
‫‪1‬‬
‫‪CE 342 – Multimedia HW# 5‬‬
‫‪H. Rabiee, Spring 2008‬‬
‫ﺷﮑﻞ‪ :1‬ﺑﻠﻮك دﯾﺎﮔﺮم ﯾﮏ ﺳﯿﺴﺘﻢ ﮐﺪﯾﻨﮓ ﻣﻌﻤﻮﻟﯽ‬
‫‪ -1-2‬ﮐﺪﯾﻨﮓ ﺑﺪون ﺗﻠﻔﺎت‬
‫‪ -1-1-2‬ﮐﺪﯾﻨﮓ ﺑﺪون ﺗﻠﻔﺎت‬
‫ﺩﺭ ﻛﺪﻳﻨﮓ ﺑﺎ ﻃﻮﻝ ﻣﺘﻐﻴﺮ )‪ (VLC‬ﺳﻤﺒﻮﻝ ﻫﺎﻳﻲ ﻛﻪ ﺍﺣﺘﻤﺎﻝ ﺣﻀﻮﺭﺷﺎﻥ ﺑﻴﺸﺘﺮ ﺍﺳﺖ ﺑﺎ ﺑﻴﺖ ﻫﺎﻱ ﻛﻤﺘﺮﻱ ﻧﻤﺎﻳﺶ ﺩﺍﺩﻩ ﻣﻲ ﺷﻮﻧﺪ )ﺑﺎ‬
‫ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻳﻚ ﻛﻠﻤﻪ ﻛﺪ ﻛﻮﺗﺎﻫﺘﺮ(‪ .‬ﻗﻀﻴﻪ ﺍﻃﻼﻋﺎﺕ ﺷﺎﻧﻮﻥ ]‪ [١‬ﺑﻴﺎﻥ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﻃﻮﻝ ﻣﻴﺎﻧﮕﻴﻦ ﺑﺮﺍﻱ ﻫﺮ ﺳﻤﺒﻮﻝ‪ ، l ،‬ﺗﻮﺳﻂ ﺁﻧﺘﺮﻭﭘﻲ ﻣﻨﺒﻊ‪،‬‬
‫‪ ،H‬ﻣﺤﺪﻭﺩ ﻣﻲ ﺷﻮﺩ‪:‬‬
‫‪ p n‬ﺍﺣﺘﻤﺎﻝ ‪-n‬ﺍﻣﻴﻦ ﺳﻤﺒﻮﻝ‪ l n ،‬ﻃﻮﻝ ﻛﻠﻤﻪ ﻛﺪ ‪ n‬ﻭ ‪ l‬ﻃﻮﻝ ﻣﻴﺎﻧﮕﻴﻦ ﻛﻠﻤﻪ ﻛﺪ ﺍﺳﺖ‪ H .‬ﺁﻧﺘﺮﻭﭘﻲ ﻣﻨﺒﻊ ﺍﺳﺖ ﻛﻪ ﺍﻃﻼﻋﺎﺕ ﻣﻴﺎﻧﮕﻴﻦ‬
‫ﻣﻮﺟﻮﺩ ﺩﺭ ﻳﻚ ﻣﻨﺒﻊ ﺗﺼﺎﺩﻓﻲ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ‪.‬‬
‫‪ -2-1-2‬ﮐﺪﯾﻨﮓ ﻫﺎﻓﻤﻦ‬
‫ﻗﻀﻴﻪ ﺷﺎﻧﻮﻥ ﺗﻨﻬﺎ ﻣﺤﺪﻭﺩﻩ ﺭﺍ ﻣﻲ ﺩﻫﺪ ﻭﻟﻲ ﺭﺍﻩ ﻋﻤﻠﻲ ﺗﺸﻜﻴﻞ ﻛﺪ ﺑﺮﺍﻱ ﺭﺳﻴﺪﻥ ﺑﻪ ﻣﺤﺪﻭﺩﻩ ﺭﺍ ﻧﻤﻲﺩﻫﺪ‪ .‬ﭼﻨﻴﻦ ﻛﺎﺭﻱ ﺑﺎ ﺭﻭﺷﻲ ﻛﻪ ﺑﻪ ﻧﺎﻡ‬
‫ﻛﺪﻳﻨﮓ ﻫﺎﻓﻤﻦ ﺷﻨﺎﺧﺘﻪ ﺷﺪﻩ ﺍﺳﺖ ﺻﻮﺭﺕ ﻣﻲ ﮔﻴﺮﺩ‪.‬‬
‫ﻣﺜﺎﻝ‪ :‬ﺗﺼﻮﻳﺮﻱ ﺭﺍ ﻛﻪ ﺑﻪ ‪ ٤‬ﺳﻄﺢ ﻛﻮﺍﻧﺘﻴﺰﻩ ﺷﺪﻩ ﺍﺳﺖ ﺩﺭ ﻧﻈﺮ ﺑﮕﻴﺮﻳﺪ‪ ٢ ، ۱ ، ۰ :‬ﻭ ‪ ، ۳‬ﻓﺮﺽ ﻛﻨﻴﺪ ﺍﺣﺘﻤﺎﻝ ﺍﻳﻦ ﺳﻄﻮﺡ ﺑﻪ ﺗﺮﺗﻴﺐ ‪،۱/۴۹‬‬
‫‪ ۳۶/۴۹ ،۴/۴۹‬ﻭ ‪ ۸/۴۹‬ﺑﺎﺷﺪ‪ .‬ﻃﺮﺡ ﻛﺪ ﻫﺎﻓﻤﻦ ﺩﺭ ﺷﻜﻞ ‪ ٢‬ﻧﺸﺎﻥ ﺩﺍﺩﻩ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬
‫ﺷﮑﻞ‪ :2‬ﻣﺜﺎﻟﯽ از ﮐﺪﯾﻨﮓ ﻫﺎﻓﻤﻦ‬
‫‪2‬‬
‫‪CE 342 – Multimedia HW# 5‬‬
‫‪H. Rabiee, Spring 2008‬‬
‫ﺩﺭ ﺍﻳﻦ ﻣﺜﺎﻝ‪ ،‬ﺩﺍﺭﻳﻢ‪:‬‬
‫ﻃﻮﻝ ﻣﺘﻮﺳﻂ‬
‫ﺁﻧﺘﺮﻭﭘﻲ ﻣﻨﺒﻊ‬
‫ﺷﻜﻞ ‪ ۳‬ﻣﺜﺎﻝ ﺩﻳﮕﺮﻱ ﺍﺯﻛﺪﻳﻨﮓ ﻫﺎﻓﻤﻦ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲﺩﻫﺪ‪:‬‬
‫ﺷﮑﻞ‪ :3‬ﻣﺜﺎل دﯾﮕﺮي از ﮐﺪﯾﻨﮓ ﻫﺎﻓﻤﻦ‬
‫‪ -3-1-2‬روش ﻫﺎي دﯾﮕﺮ ﮐﺪﯾﻨﮓ ﺑﺎ ﻃﻮل ﻣﺘﻐﯿﺮ‬
‫ﻛﺪﻳﻨﮓ ‪ [٢](Lempel, Ziv, Welsh) LZW‬ﺍﻟﮕﻮﺭﻳﺘﻢ ﻣﻮﺭﺩ ﺍﺳﺘﻔﺎﺩﻩ ﺩﺭ ﺑﺴﻴﺎﺭﻱ ﺍﺯ ﻧﺮﻡ ﺍﻓﺰﺍﺭﻫﺎﻱ ﻋﻤﻮﻣﻲ ﻓﺸﺮﺩﻩ ﺳﺎﺯﻱ ﺍﻃﻼﻋﺎﺕ‬
‫ﻣﺎﻧﻨﺪ ‪ gzip‬ﻭ ‪ pkzip‬ﺑﻮﺩﻩ ﺍﺳﺖ‪ .‬ﻳﻜﻲ ﺍﺯ ﻣﺸﻬﻮﺭﺗﺮﻳﻦ ﻓﺮﻣﺖ ﻫﺎﻱ ﻓﺎﻳﻞ ﻫﺎﻱ ﮔﺮﺍﻓﻴﻜﻲ )‪ (GIF‬ﻫﻢ ﺍﺯ ﻃﺮﺡ ﻛﺪﻳﻨﮓ ‪ LZW‬ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ‬
‫ﻛﻨﺪ‪.‬‬
‫ﺭﻭﺵ ﺷﻨﺎﺧﺘﻪ ﺷﺪﻩ ﺩﻳﮕﺮ‪ ،‬ﺭﻭﺵ ﻛﺪﻳﻨﮓ ﺭﻳﺎﺿﻲ ]‪ [٢‬ﺍﺳﺖ ﻛﻪ ﺍﺯ ﻛﺪﻳﻨﮓ ‪ LZW‬ﻭ ﻫﺎﻓﻤﻦ ﻗﺪﺭﺗﻤﻨﺪ ﺗﺮ ﺍﺳﺖ ﻭﻟﻲ ﻣﺤﺎﺳﺒﺎﺕ ﺑﻴﺸﺘﺮﻱ ﻧﻴﺎﺯ‬
‫ﺩﺍﺭﺩ‪.‬‬
‫‪ -4-1-2‬ﮐﺪﯾﻨﮓ ‪ RLC‬ﺗﺼﺎوﯾﺮ دوﺳﻄﺤﯽ‬
‫ﺩﺭ ﻛﺪﻳﻨﮓ ‪ RLC‬ﻳﻚ ﺑﻌﺪﻱ ﺗﺼﺎﻭﻳﺮ ﺩﻭﺳﻄﺤﻲ‪ ،‬ﭘﻴﻜﺴﻞ ﻫﺎ ﺍﺯ ﭼﭗ ﺑﻪ ﺭﺍﺳﺖ ﺩﺭ ﻫﺮ ﺧﻂ ﺍﺳﻜﻦ‪ ،‬ﺍﺳﻜﻦ ﻣﻲ ﺷﻮﻧﺪ‪ .‬ﻓﺮﺽ ﻛﻨﻴﺪ ﻛﻪ ﻫﺮ‬
‫ﺧﻂ‪ ،‬ﻫﻤﻴﺸﻪ ﺑﺎ ﭘﻴﻜﺴﻞ ﻫﺎﻱ ﺳﻔﻴﺪ‪ ،‬ﺷﺮﻭﻉ ﻭ ﺧﺎﺗﻤﻪ ﻣﻲ ﻳﺎﺑﺪ ﻭ ﺩﺭ ﻫﺮ ﺍﺳﻜﻦ‪ ،‬ﺑﻪ ﻧﻮﺑﺖ ﺗﻌﺪﺍﺩ ﺑﻴﺖ ﻫﺎﻱ ﭘﻴﻜﺴﻞ ﻫﺎﻱ ﺳﻔﻴﺪ ﻭ ﺳﻴﺎﻩ ﺷﻤﺮﺩﻩ‬
‫‪3‬‬
‫‪CE 342 – Multimedia HW# 5‬‬
‫‪H. Rabiee, Spring 2008‬‬
‫ﻣﻲ ﺷﻮﺩ ﻭ ﺁﺧﺮﻳﻦ ﻗﺴﻤﺖ ﭘﻴﻜﺴﻞ ﻫﺎﻱ ﺳﻔﻴﺪ ﺑﺎ ﻳﻚ ﺳﻤﺒﻮﻝ ﻣﺨﺼﻮﺹ ”‪) “EOL‬ﺍﻧﺘﻬﺎﻱ ﺧﻂ( ﺟﺎﻳﮕﺰﻳﻦ ﻣﻲ ﺷﻮﺩ‪ .‬ﻃﻮﻝ ﻗﻄﻌﺎﺕ ﺳﻔﻴﺪ‬
‫ﻭ ﺳﻴﺎﻩ ﺑﺎ ﻛﺘﺎﺏ – ﻛﺪﻫﺎﻱ ﻣﺠﺰﺍ ﻛﺪ ﻣﻲ ﺷﻮﺩ‪ .‬ﻛﺘﺎﺏ – ﻛﺪ‪ ،‬ﻣﺜﻼﹰ ﺑﺮﺍﻱ ﻗﻄﻌﺎﺕ ﺳﻔﻴﺪ ﺗﻮﺳﻂ ﺭﻭﺵ ﻛﺪﻳﻨﮓ ﻫﺎﻓﻤﻦ ﻃﺮﺍﺣﻲ ﻣﻲ ﺷﻮﺩ‪ .‬ﺑﺪﻳﻦ‬
‫ﺷﻜﻞ ﻛﻪ ﻃﻮﻝ ﻫﺮ ﻗﻄﻌﻪ )ﺷﺎﻣﻞ ‪ (EOL‬ﺑﻪ ﻋﻨﻮﺍﻥ ﺳﻤﺒﻮﻝ ﺩﺭ ﻧﻈﺮ ﮔﺮﻓﺘﻪ ﻣﻲ ﺷﻮﺩ‪ .‬ﻳﻚ ﻣﺜﺎﻝ ﻛﺪﻳﻨﮓ ‪ RLC‬ﺩﺭ ﺷﻜﻞ ‪ ٤‬ﻧﺸﺎﻥ ﺩﺍﺩﻩ ﺷﺪﻩ‬
‫ﺍﺳﺖ‪.‬‬
‫ﺷﮑﻞ‪ :4‬ﮐﺪﯾﻨﮓ ‪ Runlength‬ﺑﺮاي ﺗﺼﺎوﯾﺮ ‪bi-level‬‬
‫‪ -5-1-2‬ﮐﺪﯾﻨﮓ ‪ RLC‬دو ﺑﻌﺪي‬
‫ﻛﺪﻳﻨﮓ ‪ RLC‬ﻳﻚ ﺑﻌﺪﻱ ﺗﻨﻬﺎ ﻫﻤﺒﺴﺘﮕﻲ ﻣﻴﺎﻥ ﭘﻴﻜﺴﻞ ﻫﺎﻱ ﻳﻚ ﺧﻂ ﺭﺍ ﺟﺴﺘﺠﻮ ﻣﻲﻛﻨﺪ‪ .‬ﺩﺭ ﻛﺪﻳﻨﮓ ‪ RLC‬ﺩﻭ ﺑﻌﺪﻱ ﻳﺎ ﻛﺪﻳﻨﮓ ﺁﺩﺭﺱ‬
‫ﻧﺴﺒﻲ‪ ،‬ﻫﻤﺒﺴﺘﮕﻲ ﻣﻴﺎﻥ ﭘﻴﻜﺴﻞ ﻫﺎ ﺩﺭ ﺧﻂ ﺣﺎﺿﺮ ﻭ ﻧﻴﺰ ﺧﻂ ﻗﺒﻠﻲ ﺟﺴﺘﺠﻮ ﻣﻲﺷﻮﺩ‪ .‬ﺑﺎ ﺍﻳﻦ ﺭﻭﺵ‪ ،‬ﻭﻗﺘﻲ ﺗﻐﻴﻴﺮﻱ ﺩﺭ ﺭﻧﮓ ﺭﻭﻱ ﻣﻲﺩﻫﺪ‬
‫ﻓﺎﺻﻠﻪ ﺍﻳﻦ ﭘﻴﻜﺴﻞ ﺗﺎ ﻧﺰﺩﻳﻜﺘﺮﻳﻦ ﭘﻴﻜﺴﻠﻲ ﻛﻪ ﺩﺭ ﺁﻥ ﺗﻐﻴﻴﺮ ﺭﻭﻱ ﻣﻲﺩﻫﺪ )ﻫﻢ ﺑﻌﺪ ﻭ ﻫﻢ ﻗﺒﻞ ﺍﺯ ﺍﻳﻦ ﭘﻴﻜﺴﻞ(‪ ،‬ﺩﺭ ﺧﻂ ﻗﺒﻠﻲ ﻭ ﻫﻤﻴﻦ ﻃﻮﺭ‬
‫ﺁﺧﺮﻳﻦ ﭘﻴﻜﺴﻞ ﺩﺍﺭﺍﻱ ﺗﻐﻴﻴﺮﺍﺕ ﺩﺭ ﻫﻤﺎﻥ ﺧﻂ ﻣﺤﺎﺳﺒﻪ ﻣﻲﺷﻮﺩ ﻭ ﺁﻧﻜﻪ ﻓﺎﺻﻠﻪ ﻛﻮﺗﺎﻫﺘﺮﻱ ﺩﺍﺷﺖ‪ ،‬ﺑﻪ ﻫﻤﺮﺍﻩ ﻳﻚ ﺍﻧﺪﻳﺲ ﻛﻪ ﻧﻮﻉ ﻓﺎﺻﻠﻪ ﺭﺍ‬
‫ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ ﻛﺪ ﻣﻲﺷﻮﺩ‪ .‬ﺷﻜﻞ ‪ ٦-١٧‬ﺭﺍ ﺩﺭ ]‪ [١‬ﺑﺒﻴﻨﻴﺪ‪.‬‬
‫‪ CCITT -6-1-2‬ﮔﺮوه ‪ 3‬و ﮔﺮوه ‪ 4‬اﺳﺘﺎﻧﺪارد ﮐﺪﯾﻨﮓ دورﻧﮕﺎر‪ ،‬ﮐﺪ‪ ،1] READ‬ﻓﺼﻞ ‪(6 .6‬‬
‫ﺩﺭ ﺭﻭﺵﻫﺎﻱ ﮔﺮﻭﻩ ‪ ،٣‬ﺍﻭﻟﻴﻦ ﺧﻂ ﺩﺭ ﻫﺮ ‪ k‬ﺧﻂ ﺑﺎ ﻛﺪﻳﻨﮓ ‪ RLC‬ﻳﻚ ﺑﻌﺪﻱ ﻛﺪ ﻣﻲﺷﻮﺩ ﻭ ‪ k-1‬ﺧﻂ ﺑﺎﻗﻴﻤﺎﻧﺪﻩ ﺑﺎ ﻛﺪﻳﻨﮓ ‪٢ RLC‬‬
‫ﺑﻌﺪﻱ ﻛﻪ ﺑﺎ ﻧﺎﻡ ‪ READ1‬ﺷﻨﺎﺧﺘﻪ ﻣﻲ ﺷﻮﺩ‪ ،‬ﻛﺪ ﻣﻲﺷﻮﻧﺪ‪ .‬ﺑﺮﺍﻱ ﺟﺰﻳﻴﺎﺕ ﺍﻳﻦ ﺭﻭﺵ ﻭ ﺟﺪﺍﻭﻝ ﻛﺪ ﻭﺍﻗﻌﻲ‪ ،‬ﺑﺨﺶ ‪ [۱] ٦-٦-١‬ﺭﺍ ﺑﺒﻨﻴﺪ‪.‬‬
‫ﺩﻟﻴﻞ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ‪ RLC‬ﻳﻚ ﺑﻌﺪﻱ ﺑﺮﺍﻱ ﻫﺮ ‪ k‬ﺧﻂ‪ ،‬ﺗﻀﻌﻴﻒ ﺍﻧﺘﺸﺎﺭ ﺧﻄﺎﻱ ﺍﻧﺘﻘﺎﻝ ﺍﺳﺖ‪ .‬ﺩﺭ ﻏﻴﺮ ﺍﻳﻦ ﺻﻮﺭﺕ ﺩﺭ ﺻﻮﺭﺕ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺭﻭﺵ‬
‫‪ READ‬ﺑﻪ ﻃﻮﺭ ﭘﻴﻮﺳﺘﻪ‪ ،‬ﻭﻗﺘﻲ ﻳﻚ ﺑﻴﺖ ﺧﻄﺎ ﺟﺎﻳﻲ ﺩﺭ ﺧﻼﻝ ﺍﻧﺘﻘﺎﻝ ﺭﻭﻱ ﺑﺪﻫﺪ‪ ،‬ﺗﻤﺎﻡ ﺻﻔﺤﻪ ﺗﺄﺛﻴﺮ ﻣﻲﭘﺬﻳﺮﺩ‪.‬‬
‫ﺭﻭﺵ ﮔﺮﻭﻩ ‪ ٤‬ﺑﺮﺍﻱ ﺭﺳﺎﻧﺔ ﺍﻧﺘﻘﺎﻝ ﺍﻣﻦﺗﺮ ﻃﺮﺍﺣﻲ ﺷﺪﻩ ﺍﺳﺖ‪ ،‬ﻣﺎﻧﻨﺪ ﺧﻄﻮﻁ ﺩﺍﺩﺓ ﺍﺳﺘﻴﺠﺎﺭﻱ )‪ (leased‬ﻛﻪ ﻧﺮﺥ ﺧﻄﺎﻱ ﺑﺴﻴﺎﺭ ﻛﻮﭼﻜﻲ‬
‫ﺩﺍﺭﻧﺪ‪ .‬ﺍﻟﮕﻮﺭﻳﺘﻢ ﺷﻜﻞ ﺳﺎﺩﻩ ﺷﺪﻩﺍﻱ ﺍﺯ ﺭﻭﺵ ﮔﺮﻭﻩ ‪ ۳‬ﺍﺳﺖ ﻛﻪ ‪ RLC‬ﻳﻚ ﺑﻌﺪﻱ ﺣﺬﻑ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬
‫‪Relative Element Address Designate‬‬
‫‪4‬‬
‫‪CE 342 – Multimedia HW# 5‬‬
‫‪H. Rabiee, Spring 2008‬‬
‫‪1‬‬
‫‪ -2-2‬ﮐﺪﯾﻨﮓ ﭘﯿﺶ ﺑﯿﻨﯽ ﮐﻨﻨﺪه‬
‫ﻣﻘﺪﺍﺭ ﭘﻴﻜﺴﻞ ﺣﺎﺿﺮ ﻣﻌﻤﻮﻻﹰ ﺗﻐﻴﻴﺮﺍﺕ ﺳﺮﻳﻌﻲ ﻧﺴﺒﺖ ﺑﻪ ﭘﻴﻜﺴﻞ ﻫﺎﻱ ﻣﺠﺎﻭﺭ ﻧﺪﺍﺭﺩ‪ .‬ﺑﻨﺎﺑﺮﺍﻳﻦ ﻣﻲﺗﻮﺍﻧﺪ ﺍﺯ ﺭﻭﻱ ﻧﻤﻮﻧﻪ ﻫﺎﻱ ﻗﺒﻠﻲ ﭘﻴﺶﺑﻴﻨﻲ‬
‫ﺷﻮﺩ‪ .‬ﺑﻨﺎﺑﺮﺍﻳﻦ‪ ،‬ﺑﻪ ﺟﺎﻱ ﻛﺪ ﻛﺮﺩﻥ ﭘﻴﻜﺴﻞ ﻫﺎﻱ ﺍﺻﻠﻲ‪ ،‬ﻣﻲﺗﻮﺍﻥ ﭘﻴﻜﺴﻞ ﺣﺎﺿﺮ ﺭﺍ ﺍﺯ ﺭﻭﻱ ﻗﺒﻠﻲ ﻫﺎ ﭘﻴﺶ ﺑﻴﻨﻲ ﻛﺮﺩ ﻭ ﺗﻨﻬﺎ ﺧﻄﺎﻱ ﭘﻴﺶﺑﻴﻨﻲ ﺭﺍ‬
‫ﻣﺸﺨﺺ ﻛﺮﺩ‪ .‬ﺍﻳﺪﺓ ﻛﺪﻳﻨﮓ ﺑﺎ ﭘﻴﺶ ﺑﻴﻨﻲ ﻗﺒﻼﹰ ﺩﺭ ﻛﺪﻳﻨﮓ ﺻﺤﺒﺖ ﻭ ﺻﺪﺍ ﺍﺳﺘﻔﺎﺩﻩ ﺷﺪﻩ ﺍﺳﺖ‪ .‬ﻛﺪﻳﻨﮓ ﭘﻴﺶ ﺑﻴﻨﻲ ﻛﻨﻨﺪﻩ ﻣﻲﺗﻮﺍﻧﻨﺪ ﺑﺎ ﺗﻠﻔﺎﺕ‬
‫ﻳﺎ ﺑﺪﻭﻥ ﺗﻠﻔﺎﺕ ﺑﺎﺷﺪ‪ ،‬ﺩﺭ ﻛﺪﻳﻨﮓ ﭘﻴﺶ ﺑﻴﻨﻲ ﻛﻨﻨﺪﻩ ﺑﺎ ﺗﻠﻔﺎﺕ‪ ،‬ﺧﻄﺎﻱ ﭘﻴﺶﺑﻴﻨﻲ ﺍﺑﺘﺪﺍ ﻛﻮﺍﻧﺘﻴﺰﻩ ﺷﺪﻩ ﻭ ﺳﭙﺲ ﻛﺪ ﻣﻲﺷﻮﺩ‪ .‬ﺩﻟﻴﻞ ﺍﻳﻦ ﻛﺎﺭ ﺍﻳﻦ‬
‫ﺍﺳﺖ ﻛﻪ ﭘﻴﺶﺑﻴﻨﻲ ﺑﻪ ﺷﻜﻠﻲ ﻛﻪ ﺩﺭ ﺍﺩﺍﻣﻪ ﺧﻮﺍﻫﻴﻢ ﮔﻔﺖ ﺑﻪ ﻛﺎﻫﺶ ﻧﺮﺥ ﺑﻴﺖ ﻛﻤﻚ ﻣﻲﻛﻨﺪ‪ .‬ﺧﻄﺎﻱ ﭘﻴﺶ ﺑﻴﻨﻲ ﺗﻮﺯﻳﻊ ﻏﻴﺮ ﻳﻜﻨﻮﺍﺧﺖ ﺧﻮﺍﻫﺪ‬
‫ﺩﺍﺷﺖ ﻛﻪ ﻋﻤﺪﺗﺎﹰ ﺩﺭ ﻧﺰﺩﻳﻜﻲ ﺻﻔﺮ ﻣﺘﻤﺮﻛﺰ ﺍﺳﺖ ﻭ ﺁﻧﺘﺮﻭﭘﻲ ﻛﻤﺘﺮﻱ ﻧﺴﺒﺖ ﺑﻪ ﻧﻤﻮﻧﻪﻫﺎﻱ ﺍﺻﻠﻲ ﺩﺍﺭﺩ ﻛﻪ ﻣﻌﻤﻮﻻﹰ ﺗﻮﺯﻳﻊ ﻳﻜﻨﻮﺍﺧﺖ ﺩﺍﺭﻧﺪ‪ .‬ﺑﺎ‬
‫ﻛﺪﻳﻨﮓ ﺁﻧﺘﺮﻭﭘﻲ )ﻣﺜﻞ ﻛﺪﻳﻨﮓ ﻫﺎﻓﻤﻦ( ﻣﻘﺎﺩﻳﺮ ﺧﻄﺎ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺑﺎ ﺑﻴﺖ ﻫﺎﻱ ﻛﻤﺘﺮﻱ ﻧﺴﺒﺖ ﺑﻪ ﻣﻘﺎﺩﻳﺮ ﻧﻤﻮﻧﻪ ﻫﺎﻱ ﺗﺼﻮﻳﺮ ﺍﺻﻠﻲ ﻛﺪ ﺷﻮﻧﺪ‪.‬‬
‫ﺷﻜﻞ ‪ ٥‬ﺑﻠﻮﻙ ﺩﻳﺎﮔﺮﺍﻡ ﻛﻠﻲ ﺑﺮﺍﻱ ﻛﺪﮔﺬﺍﺭ ﻭ ﻛﺪ ﮔﺸﺎﻱ ﻳﻚ ﺳﻴﺴﺘﻢ ﻛﺪﻳﻨﮓ ﺑﺎ ﺗﻠﻔﺎﺕ ﭘﻴﺶ ﺑﻴﻨﻲ ﻛﻨﻨﺪﻩ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ‪ .‬ﺩﺭ ﻳﻚ ﻛﺪ ﻛﻨﻨﺪﻩ‬
‫ﺑﺪﻭﻥ ﺗﻠﻔﺎﺕ‪ ،‬ﻣﺮﺣﻠﻪ ﻛﻮﺍﻧﺘﻴﺰﺍﺳﻴﻮﻥ ﺍﻧﺠﺎﻡ ﻧﻤﻲﺷﻮﺩ‪ .‬ﺷﻜﻞ ‪ ٦‬ﭼﮕﻮﻧﮕﻲ ﺍﺳﺘﻔﺎﺩﺓ ﻛﺪﻳﻨﮓ ﭘﻴﺶ ﺑﻴﻨﻲ ﻛﻨﻨﺪﻩ ﺑﺮﺍﻱ ﻳﻚ ﺗﺼﻮﻳﺮ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲﺩﻫﺪ‪.‬‬
‫ﺷﮑﻞ‪ :5‬ﺑﻠﻮك دﯾﺎﮔﺮام ﮐﻠﯽ ﺑﺮاي ﮐﺪﮔﺬار و ﮐﺪﮔﺸﺎي ﯾﮏ ﺳﯿﺴﺘﻢ ﮐﺪﯾﻨﮓ ﺑﺎ ﺗﻠﻔﺎت ﭘﯿﺶ ﺑﯿﻨﯽ ﮐﻨﻨﺪه‬
‫ﺷﮑﻞ‪ :6‬ﮐﺪﯾﻨﮓ ﭘﯿﺶ ﺑﯿﻨﯽ ﺷﺪه‬
‫‪ -3-2‬ﮐﺪﯾﻨﮓ ﺗﺒﺪﯾﻠﯽ )ﮐﺪﯾﻨﮓ ﺑﺎ ﺗﻠﻔﺎت( ]‪ ،1‬ﻓﺼﻞ ‪[6-5‬‬
‫ﻛﺪﻳﻨﮓ ﺑﺪﻭﻥ ﺗﻠﻔﺎﺕ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﻪ ﻧﺴﺒﺖ ﻫﺎﻱ ﻓﺸﺮﺩﻩ ﺳﺎﺯﻱ ‪ ٢‬ﺗﺎ ‪ ٣‬ﺑﺮﺍﻱ ﺑﻴﺸﺘﺮ ﺗﺼﺎﻭﻳﺮ ﺑﺮﺳﺪ‪ .‬ﺑﺮﺍﻱ ﻛﺎﻫﺶ ﺑﻴﺸﺘﺮ ﻣﻘﺪﺍﺭ ﺩﺍﺩﻩﻫﺎ‪ ،‬ﺭﻭﺵ‬
‫ﻫﺎﻱ ﺑﺎ ﺗﻠﻔﺎﺕ ﻛﺪﻳﻨﮓ‪ ،‬ﻛﻮﺍﻧﺘﻴﺰﺍﺳﻴﻮﻥ ﺭﺍ ﺑﺮ ﺭﻭﻱ ﻣﻘﺎﺩﻳﺮ ﺗﺼﺎﻭﻳﺮ ﺍﺻﻠﻲ ﻳﺎ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻣﻴﺪﺍﻥ ﺗﺒﺪﻳﻞ ﺳﻴﮕﻨﺎﻝ ﺍﺻﻠﻲ ﺑﻪ ﻛﺎﺭ ﻣﻲﺑﺮﻧﺪ‪ .‬ﺗﺒﺪﻳﻞ ﺍﺯ‬
‫ﻫﻤﺒﺴﺘﮕﻲ ﺁﻣﺎﺭﻱ ﻣﻴﺎﻥ ﻧﻤﻮﻧﻪﻫﺎﻱ ﺍﺻﻠﻲ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲﻛﻨﺪ‪ .‬ﺭﻭﺵﻫﺎﻱ ﻣﺤﺒﻮﺏ‪ ،‬ﺷﺎﻣﻞ ﭘﻴﺶ ﺑﻴﻨﻲ ﺧﻄﻲ ﻭ ﺗﺒﺪﻳﻞ ﻫﺎﻱ ﻳﻚ ﺑﻪ ﻳﻚ ﻣﻲﺑﺎﺷﺪ‪.‬‬
‫ﺩﺭ ﺍﻳﻦ ﺑﺨﺶ ﺑﺮ ﺭﻭﻱ ﻛﺪﻳﻨﮓ ﺗﺒﺪﻳﻠﻲ ﻣﺘﻤﺮﻛﺰ ﻣﻲﺷﻮﻳﻢ ﻛﻪ ﺑﺮﺍﻱ ﺗﺼﺎﻭﻳﺮ‪ ،‬ﻛﺎﺭﺍﻳﻲ ﺑﻴﺸﺘﺮﻱ ﺩﺍﺭﺩ‪.‬‬
‫‪5‬‬
‫‪CE 342 – Multimedia HW# 5‬‬
‫‪H. Rabiee, Spring 2008‬‬
‫ﻳﻜﻲ ﺍﺯ ﻣﺤﺒﻮﺑﺘﺮﻳﻦ ﺭﺍﻫﻜﺎﺭﻫﺎﻱ ﻛﺪﻳﻨﮓ ﺑﺎ ﺗﻠﻔﺎﺕ ﺑﺮﺍﻱ ﺗﺼﺎﻭﻳﺮ‪ ،‬ﻛﺪﻳﻨﮓ ﺗﺒﺪﻳﻠﻲ ﺍﺳﺖ‪ .‬ﺩﺭ ﻛﺪﻳﻨﮓ ﺗﺒﺪﻳﻠﻲ ﻣﺒﺘﻨﻲ ﺑﺮ ﺑﻠﻮﻙ‪ ،‬ﺗﺼﻮﻳﺮ ﺑﻪ‬
‫ﺑﻠﻮﻙﻫﺎﻱ ﺑﺪﻭﻥ ﻫﻤﭙﻮﺷﺎﻧﻲ ﺗﻘﺴﻴﻢ ﻣﻲ ﺷﻮﺩ‪ .‬ﻫﺮ ﺑﻠﻮﻙ ﺑﺎ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺍﻟﮕﻮﻫﺎﻱ ﭘﺎﻳﻪ ﻧﺸﺎﻥ ﺩﺍﺩﻩ ﻣﻲ ﺷﻮﺩ‪) .‬ﻛﻪ ﺑﻪ ﻋﻨﻮﺍﻥ ﺗﻮﺍﺑﻊ ﭘﺎﻳﻪ ﺗﺒﺪﻳﻞ‬
‫ﺷﻨﺎﺧﺘﻪ ﻣﻲﺷﻮﻧﺪ( )ﺷﻜﻞ ‪ (٧‬ﺳﻬﻢ ﻫﺮ ﺍﻟﮕﻮﻱ ﭘﺎﻳﻪ ﺑﻪ ﻋﻨﻮﺍﻥ ﺿﺮﻳﺐ ﺗﺒﺪﻳﻞ ﺷﻨﺎﺧﺘﻪ ﻣﻲﺷﻮﺩ‪ .‬ﺍﺯ ﻧﻈﺮ ﺭﻳﺎﺿﻲ ﺍﻳﻦ ﺿﺮﺍﻳﺐ ﺑﺎ ﻋﻤﻠﻴﺎﺕ‬
‫ﻣﺎﺗﺮﻳﺴﻲ ﺑﻪ ﺩﺳﺖ ﻣﻲﺁﻳﻨﺪ‪ .‬ﻣﻘﺎﺩﻳﺮ ﭘﻴﻜﺴﻞﻫﺎﻱ ﺍﺻﻠﻲ ﺩﺍﺧﻞ ﻫﺮ ﺑﻠﻮﻙ‪ ،‬ﻳﻚ ﺑﺮﺩﺍﺭ ﺭﺍ ﺗﺸﻜﻴﻞ ﻣﻲ ﺩﻫﻨﺪ‪ .‬ﺩﺭ ﺣﺎﻟﻲ ﻛﻪ ﺗﻤﺎﻡ ﺗﻮﺍﺑﻊ ﭘﺎﻳﻪ ﺗﺒﺪﻳﻞ‪،‬‬
‫ﻳﻚ ﻣﺎﺗﺮﻳﺲ ﺭﺍ ﻣﻲﺳﺎﺯﻧﺪ‪ .‬ﺑﺮﺩﺍﺭ ﺷﺎﻣﻞ ﺗﻤﺎﻡ ﺿﺮﺍﻳﺐ ﺗﺒﺪﻳﻞ‪ ،‬ﺣﺎﺻﻠﻀﺮﺏ ﻣﺎﺗﺮﻳﺲ ﺗﺒﺪﻳﻞ ﺑﺎ ﺑﺮﺩﺍﺭ ﭘﻴﻜﺴﻞﻫﺎ ﺍﺳﺖ‪ .‬ﺗﺒﺪﻳﻞ ﻃﻮﺭﻱ ﻃﺮﺍﺣﻲ‬
‫ﺷﺪﻩ ﺍﺳﺖ ﻛﻪ ﺍﻧﺮﮊﻱ ﺳﻴﮕﻨﺎﻝ ﺍﺻﻠﻲ ﺗﻨﻬﺎ ﺩﺭ ﺿﺮﺍﻳﺐ ﻛﻤﻲ ﺟﻤﻊ ﺷﻮﺩ ﻭ ﻫﻤﺒﺴﺘﮕﻲ ﻣﻴﺎﻥ ﻣﺘﻐﻴﺮﻫﺎ ﺑﺮﺍﻱ ﻛﺪ ﺷﺪﻥ ﻛﺎﻫﺶ ﻳﺎﺑﺪ‪ .‬ﻫﺮ ﺩﻭﻱ ﺍﻳﻦ‬
‫ﻣﻮﺍﺭﺩ ﺑﻪ ﻛﺎﻫﺶ ﻧﺮﺥ ﺑﻴﺖ ﻛﻤﻚ ﻣﻲﻛﻨﻨﺪ‪ .‬ﺷﻜﻞ ‪ ،٧‬ﺑﻮﺳﻴﻠﻪ ﺗﺒﺪﻳﻞ ﺑﻠﻮﻛﻲ‪ ،‬ﻫﺮ ﺑﻠﻮﻙ ﺗﺼﻮﻳﺮ ﺭﺍ ﺑﻪ ﺻﻮﺭﺕ ﺗﺮﻛﻴﺐ ﺧﻄﻲ ﺑﻌﻀﻲ ﺍﻟﮕﻮﻫﺎﻱ‬
‫ﭘﺎﻳﻪ ﻧﺸﺎﻥ ﻣﻲﺩﻫﺪ‪.‬‬
‫ﺷﮑﻞ ‪ :7‬ﺗﺮﮐﯿﺐ ﺧﻄﯽ ﺑﻌﻀﯽ اﻟﮕﻮﻫﺎي ﭘﺎﯾﻪ‬
‫ﺷﻜﻞ ‪ ٨‬ﭘﺮﺩﺍﺯﺵ ﻳﻚ ﺑﻠﻮﻙ ﻛﺪﻳﻨﮕﻲ ﺗﺒﺪﻳﻠﻲ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ ﺍﺑﺘﺪﺍ ﺑﻠﻮﻙ ﻭﺭﻭﺩﻱ ﺍﺯ ﻃﺮﻳﻖ ﻳﻚ ﻋﻤﻠﮕﺮ ﻣﺎﺗﺮﻳﺴﻲ ﺗﺒﺪﻳﻞ ﻣﻲﺷﻮﺩ‪ .‬ﺿﺮﺍﻳﺐ‬
‫ﺗﺒﺪﻳﻞ‪ ،‬ﺳﭙﺲ ﻛﻮﺍﻧﺘﻴﺰﻩ ﺷﺪﻩ ﻭ ﻛﺪ ﻣﻲﺷﻮﻧﺪ‪ .‬ﺩﺭ ﻛﺪﮔﺸﺎ‪ ،‬ﺑﻠﻮﻙ ﺍﺻﻠﻲ ﺍﺯ ﺭﻭﻱ ﺿﺮﺍﻳﺐ ﻛﻮﺍﻧﺘﻴﺰﻩ ﺷﺪﻩ ﺍﺯ ﻃﺮﻳﻖ ﻋﻜﺲ ﺗﺒﺪﻳﻞ ﺑﺎﺯﺳﺎﺯﻱ‬
‫ﻣﻲﺷﻮﺩ‪.‬‬
‫ﺷﮑﻞ ‪ :8‬ﭘﺮدازش ﯾﮏ ﺑﻠﻮك ﮐﺪﯾﻨﮕﯽ ﺗﺒﺪﯾﻠﯽ‬
‫‪ -1-3-2‬ﺗﺒﺪﯾﻞ ﮐﺴﯿﻨﻮﺳﯽ ﮔﺴﺴﺘﻪ )‪(DCT‬‬
‫‪ DCT‬ﺑﺮﺍﻱ ﺳﻴﮕﻨﺎﻝ ﻫﺎﻱ ﺗﺼﻮﻳﺮ‪ ،‬ﻣﺤﺒﻮﺏ ﺍﺳﺖ‪ ،‬ﺯﻳﺮﺍ ﺑﻪ ﺧﻮﺑﻲ ﺑﺎ ﺧﺼﻮﺻﻴﺎﺕ ﺁﻣﺎﺭﻱ ﺳﻴﮕﻨﺎﻝﻫﺎﻱ ﺗﺼﻮﻳﺮ ﻣﻮﺟﻮﺩ ﻣﻨﻄﺒﻖ ﺍﺳﺖ‪.‬‬
‫ﺑﺮﺩﺍﺭﻫﺎﻱ ﭘﺎﻳﻪ ﻳﻚ ﺑﻌﺪﻱ ‪-N‬ﻧﻘﻄﻪ ﺍﻱ ‪ DCT‬ﺑﺪﻳﻦ ﺻﻮﺭﺕ ﺗﻌﺮﻳﻒ ﻣﻲ ﺷﻮﻧﺪ‪:‬‬
‫ﺗﺒﺪﻳﻞ ﻭ ﻋﻜﺲ ﺗﺒﺪﻳﻞ ﺑﺪﻳﻦ ﺻﻮﺭﺕ ﺗﻌﺮﻳﻒ ﻣﻲ ﺷﻮﻧﺪ‪:‬‬
‫‪6‬‬
‫‪CE 342 – Multimedia HW# 5‬‬
‫‪H. Rabiee, Spring 2008‬‬
‫ﺗﻮﺟﻪ ﻛﻨﻴﺪ ﻛﻪ ﺑﺮﺩﺍﺭﻫﺎﻱ ﭘﺎﻳﻪ ﺩﺭ ﺍﻟﮕﻮﻱ ﺳﻴﻨﻮﺳﻲ ﺑﺎ ﺍﻓﺰﺍﻳﺶ ﻓﺮﻛﺎﻧﺲ‪ ،‬ﺗﻐﻴﻴﺮ ﻣﻲﻛﻨﻨﺪ‪ .‬ﺗﻮﺟﻪ ﻛﻨﻴﺪ ﻛﻪ ‪-N DCT‬ﻧﻘﻄﻪﺍﻱ ﺑﻪ ‪-٢N DFT‬‬
‫ﻧﻘﻄﻪﺍﻱ ﻣﺮﺗﺒﻂ ﺍﺳﺖ ﻭﻟﻲ ﺟﺰﺀ ﺣﻘﻴﻘﻲ ﺁﻥ ﻧﻴﺴﺖ‪ .‬ﻫﺮ ﺿﺮﻳﺐ ‪ DCT‬ﺳﻬﻢ ﺍﻟﮕﻮﻱ ﺳﻴﻨﻮﺳﻲ ﺭﺍ ﺩﺭ ﻳﻚ ﻓﺮﻛﺎﻧﺲ ﻣﺸﺨﺺ ﺩﺭ ﺳﻴﮕﻨﺎﻝ ﺍﺻﻠﻲ‬
‫ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ‪ .‬ﻛﻮﭼﻜﺘﺮﻳﻦ ﺿﺮﻳﺐ‪ ،‬ﻛﻪ ﺑﻪ ﻋﻨﻮﺍﻥ ﺿﺮﻳﺐ ‪ DC‬ﺷﻨﺎﺧﺘﻪ ﻣﻲ ﺷﻮﺩ‪ ،‬ﻣﻴﺎﻧﮕﻴﻦ ﻣﻘﺎﺩﻳﺮ ﺳﻴﮕﻨﺎﻝ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲﺩﻫﺪ‪ .‬ﺑﻘﻴﻪ ﺿﺮﺍﻳﺐ ﻛﻪ‬
‫ﺑﻪ ﻋﻨﻮﺍﻥ ﺿﺮﺍﻳﺐ ‪ AC‬ﺷﻨﺎﺧﺘﻪ ﻣﻲ ﺷﻮﻧﺪ‪ ،‬ﺑﻪ ﻓﺮﻛﺎﻧﺲﻫﺎﻱ ﺑﺎﻻﺗﺮ ﻣﺮﺗﺒﻂ ﻫﺴﺘﻨﺪ‪.‬‬
‫ﺑﺮﺍﻱ ﺑﻪ ﺩﺳﺖ ﺁﻭﺭﺩﻥ ‪ ٢ DCT‬ﺑﻌﺪﻱ ﺍﺯ ﻳﻚ ﺑﻠﻮﻙ ﺗﺼﻮﻳﺮ‪ ،‬ﻓﺮﺩ ﻣﻲ ﺗﻮﺍﻧﺪ ﺩﺭ ﺍﺑﺘﺪﺍ ‪ DCT‬ﻳﻚ ﺑﻌﺪﻱ ﺭﺍ ﺑﺮ ﺭﻭﻱ ﻫﺮ ﺭﺩﻳﻒ ﺑﻠﻮﻙ ﺗﺼﻮﻳﺮ‬
‫ﺑﻪ ﻛﺎﺭ ﺑﺮﺩ ﻭ ﺳﭙﺲ ‪ DCT‬ﻳﻚ ﺑﻌﺪﻱ ﺭﺍ ﺑﻪ ﺭﻭﻱ ﻫﺮ ﺳﺘﻮﻥ ﺑﻠﻮﻙ ﺣﺎﺻﻞ ﺍﻧﺠﺎﻡ ﺩﻫﺪ‪ .‬ﻳﻚ ﺗﺒﺪﻳﻞ ‪ ٢‬ﺑﻌﺪﻱ ﻣﻌﺎﺩﻝ ﻧﻤﺎﻳﺶ ﺳﻴﮕﻨﺎﻝ ﺑﻮﺳﻴﻠﻪ‬
‫ﺟﻤﻊ ﺗﻌﺪﺍﺩ ﺯﻳﺎﺩﻱ ﺍﻟﮕﻮﻫﺎﻱ ﭘﺎﻳﻪ ﺍﺳﺖ‪ .‬ﺍﻟﮕﻮﻫﺎﻱ ﺑﻠﻮﻙ ﭘﺎﻳﻪ ﻣﻨﺘﺎﻇﺮ ﺑﺎ ‪ ٨*٨ DCT‬ﺩﺭ ﺷﻜﻞ ‪ ٩‬ﻧﺸﺎﻥ ﺩﺍﺩﻩ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬
‫ﺍﻳﻦ ﺷﻜﻞ ﺑﺎ ﺩﺳﺘﻮﺭﺍﺕ ‪ MATLAB‬ﺯﻳﺮ ﺑﻪ ﺩﺳﺖ ﺁﻣﺪﻩ ﺍﺳﺖ‪.‬‬
‫;)‪T=dctmtx2(8‬‬
‫;‪figure; colormap(gray(256));n=1‬‬
‫)‪for (i=1:8‬‬
‫)‪for (j=1:8‬‬
‫‪subplot(8,8,n),‬‬
‫;))‪imagesc(T(i,:)'*T(j,:‬‬
‫;‪axis off; axis image‬‬
‫;‪n=n+1‬‬
‫;‪end‬‬
‫;‪end‬‬
‫ﺷﮑﻞ ‪ :9‬اﻟﮕﻮﻫﺎي ﺑﻠﻮك ﭘﺎﯾﻪ ﻣﻨﺘﺎﻇﺮ ﺑﺎ ‪8*8 DCT‬‬
‫‪7‬‬
‫‪CE 342 – Multimedia HW# 5‬‬
‫‪H. Rabiee, Spring 2008‬‬
‫‪ -2-3-2‬ﻧﻤﺎﯾﺶ ﺑﻠﻮك ﻫﺎي ﺗﺼﻮﯾﺮ ﺑﺎ ‪DCT‬‬
‫ﺩﻟﻴﻞ ﻣﻨﺎﺳﺐ ﺑﻮﺩﻥ ‪ DCT‬ﺑﺮﺍﻱ ﻓﺸﺮﺩﻩ ﺳﺎﺯﻱ ﺗﺼﻮﻳﺮ ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ﻳﻚ ﺑﻠﻮﻙ ﺗﺼﻮﻳﺮ ﻣﻌﻤﻮﻻﹰ ﻣﻲﺗﻮﺍﻧﺪ ﺗﻮﺳﻂ ﺿﺮﺍﻳﺐ ‪ DCT‬ﺑﺴﻴﺎﺭ‬
‫ﻓﺮﻛﺎﻧﺲ ﭘﺎﻳﻴﻦ ﻧﺸﺎﻥ ﺩﺍﺩﻩ ﺷﻮﺩ‪.‬‬
‫ﺍﻳﻦ ﻛﺎﺭ ﺑﻪ ﺍﻳﻦ ﺧﺎﻃﺮ ﺍﺳﺖ ﻛﻪ ﻣﻘﺎﺩﻳﺮ ﺷﺪﺕ ﺩﺭ ﻫﺮ ﺗﺼﻮﻳﺮ ﻣﻌﻤﻮﻻﹰ ﺑﻪ ﺻﻮﺭﺕ ﻫﻤﻮﺍﺭ ﺗﻐﻴﻴﺮ ﻣﻲ ﻛﻨﻨﺪ ﻭ ﻣﺆﻟﻔﻪﻫﺎﻱ ﻓﺮﻛﺎﻧﺲ ﺧﻴﻠﻲ ﺑﺎﻻ ﺗﻨﻬﺎ‬
‫ﺩﺭ ﻧﺰﺩﻳﻜﻲ ﻟﺒﻪ ﻫﺎ ﻭﺟﻮﺩ ﺩﺍﺭﻧﺪ‪ .‬ﺩﺭ ﺷﻜﻞ ‪ ١٠‬ﺗﻘﺮﻳﺐ ﻳﻚ ﺗﺼﻮﻳﺮ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺗﻌﺪﺍﺩ ﻣﺨﺘﻠﻔﻲ ﺍﺯ ﺿﺮﺍﻳﺐ ‪ DCT‬ﻧﺸﺎﻥ ﺩﺍﺩﻩ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬
‫ﻣﻲﺗﻮﺍﻧﻴﻢ ﺑﺒﻴﻨﻴﻢ ﻛﻪ ﻓﻘﻂ ‪ ١٦‬ﺗﺎ ﺍﺯ ‪ ٦٤‬ﺿﺮﻳﺐ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﻪ ﺧﻮﺑﻲ ﺑﻠﻮﻙ ﺍﺻﻠﻲ ﺭﺍ ﻧﻤﺎﻳﺶ ﺩﻫﺪ‪ ،‬ﺷﻤﺎ ﻣﻲﺗﻮﺍﻧﻴﺪ ﺩﻗﺖ ﺗﻘﺮﻳﺐ ﺿﺮﺍﻳﺐ ﺑﺎ ﺗﻌﺪﺍﺩ‬
‫ﻣﺨﺘﻠﻒ ‪ DCT‬ﺭﺍ ﺑﺎ ﺑﺮﻧﺎﻣﺔ ‪ dctdemo ،MATLAB‬ﺁﺯﻣﺎﻳﺶ ﻛﻨﻴﺪ‪.‬‬
‫ﺷﮑﻞ ‪ :10‬ﺗﻘﺮﯾﺐ ﯾﮏ ﺗﺼﻮﯾﺮ ﺑﺎ اﺳﺘﻔﺎده از ﺗﻌﺪاد ﻣﺨﺘﻠﻔﯽ از ﺿﺮاﯾﺐ ‪ .DCT‬ﭼﭗ‪-‬ﺑﺎﻻ‪ :‬ﺷﮑﻞ اﺻﻠﯽ ﯾﺎ ﺗﻤﺎﻣﯽ ﺿﺮاﯾﺐ ‪،DCT‬‬
‫راﺳﺖ‪-‬ﺑﺎﻻ‪ 16/64 :‬ﺿﺮﯾﺐ‪ ،‬ﭼﭗ‪-‬ﭘﺎﯾﯿﻦ‪ 8/64 :‬ﺿﺮﯾﺐ‪ ،‬راﺳﺖ‪-‬ﭘﺎﯾﯿﻦ‪ 4/64 :‬ﺿﺮﯾﺐ‪.‬‬
‫‪ -3-3-2‬اﺳﺘﺎﻧﺪارد ‪ JPEG‬ﺑﺮاي ﻓﺸﺮده ﺳﺎزي ﺗﺼﻮﯾﺮ ﺛﺎﺑﺖ‬
‫ﺍﺳﺘﺎﻧﺪﺍﺭﺩ ‪ JPEG‬ﺑﺮﺍﻱ ﻓﺸﺮﺩﻩﺳﺎﺯﻱ ﺗﺼﻮﻳﺮ ﺛﺎﺑﺖ ﺗﻮﺻﻴﻪ ﺷﺪﻩ ﺗﻮﺳﻂ ﮔﺮﻭﻩ ﻣﺘﺨﺼﺼﻴﻦ ﻋﻜﺲ ‪) ISO‬ﺳﺎﺯﻣﺎﻥ ﺍﺳﺘﺎﻧﺪﺍﺭﺩﻫﺎﻱ ﺑﻴﻦﺍﻟﻤﻠﻠﻲ(‬
‫ﺍﺯ ﺳﻪ ﺑﺨﺶ ﺗﺸﻜﻴﻞ ﻳﺎﻓﺘﻪ ﺍﺳﺖ‪:‬‬
‫‪ -١‬ﺭﻭﺵ ﻓﺸﺮﺩﻩ ﺳﺎﺯﻱ ﺑﺎ ﺗﻠﻔﺎﺕ ﻣﺒﺘﻨﻲ ﺑﺮ ‪ DCT‬ﺑﺮﺍﻱ ﺩﻗﺖ ﻭ ﺭﺯﻟﻮﺷﻦ ﺍﺳﺘﺎﻧﺪﺍﺭﺩ )‪٨bits/color/pixel‬‬
‫(‪.‬‬
‫‪ -٢‬ﺭﻭﺵ ﻛﺪﻳﻨﮓ ﺗﻮﺳﻌﻪ ﻳﺎﻓﺘﻪ ﺑﺮﺍﻱ ﺗﺼﺎﻭﻳﺮ ﺑﺎ ﺩﻗﺖ ﻭ ﺭﺯﻟﻮﺷﻦ ﺑﺎﻻﺗﺮ ﻛﻪ ﺍﻟﮕﻮﺭﻳﺘﻢ ﻗﺒﻠﻲ ﺭﺍ ﺩﺭ ﺩﺍﺧﻞ ﺧﻮﺩ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲﻛﻨﺪ‪.‬‬
‫‪ -۳‬ﺭﻭﺵ ﻛﺪﻳﻨﮓ ﭘﻴﺶ ﺑﻴﻨﻲ ﻛﻨﻨﺪﻩ ﺑﺪﻭﻥ ﺗﻠﻔﺎﺕ‪.‬‬
‫ﺩﺭ ﺍﻳﻨﺠﺎ ﻣﺎ ﺗﻨﻬﺎ ﺍﻟﮕﻮﺭﻳﺘﻢ ﺍﻭﻝ ﺭﺍ ﻛﻪ ﺑﻴﺸﺘﺮ ﺍﺯ ﺑﻘﻴﻪ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲﺷﻮﺩ ﺗﻮﺿﻴﺢ ﻣﻲﺩﻫﻴﻢ‪ .‬ﺩﺭ ﺍﺑﺘﺪﺍ ﺗﺼﻮﻳﺮ ﻭﺭﻭﺩﻱ ﺭﺍ ﺑﻪ ﺑﻠﻮﻙ ﻫﺎﻱ ‪ ٨*٨‬ﺑﺪﻭﻥ‬
‫ﻫﻤﭙﻮﺷﺎﻧﻲ ﺗﻘﺴﻴﻢ ﻣﻲﻛﻨﻴﻢ‪ .‬ﺳﭙﺲ ﻳﻚ ‪ ٨*٨ DCT‬ﺑﺮ ﺭﻭﻱ ﻫﺮ ﺑﻠﻮﻙ ﺑﻪ ﻛﺎﺭ ﻣﻲﺭﻭﺩ‪ .‬ﺑﺮﺍﻱ ﺿﺮﻳﺐ ‪ DC‬ﻫﺮ ﺑﻠﻮﻙ‪ ،‬ﺭﻭﺵ ﻛﺪﻳﻨﮓ ﭘﻴﺶ‬
‫ﺑﻴﻨﻲ ﻛﻨﻨﺪﻩ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲﺷﻮﺩ‪ ،‬ﻳﻌﻨﻲ ﻣﻘﺪﺍﺭ ‪ DC‬ﺣﺎﺿﺮ ﺍﺯ ﺭﻭﻱ ﻣﻘﺪﺍﺭ ‪ DC‬ﺑﻠﻮﻙ ﻗﺒﻠﻲ ﭘﻴﺶ ﺑﻴﻨﻲ ﻣﻲﺷﻮﺩ ﻭ ﺧﻄﺎﻱ ﭘﻴﺶ ﺑﻴﻨﻲ ﺑﺎ ﻳﻚ‬
‫ﻛﻮﺍﻧﺘﻴﺰﻩ ﻛﻨﻨﺪﺓ ﻳﻜﻨﻮﺍﺧﺖ ﻛﻮﺍﻧﺘﻴﺰﻩ ﻣﻲﺷﻮﺩ‪ .‬ﺿﺮﺍﻳﺐ ‪ AC‬ﺩﻳﮕﺮ ﻣﺴﺘﻘﻴﻤﺎﹰ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻛﻮﺍﻧﺘﺎﻳﺰﺭﻫﺎﻱ ﻣﺨﺘﻠﻒ‪ ،‬ﻛﻮﺍﻧﺘﻴﺰﻩ ﻣﻲ ﺷﻮﻧﺪ )ﺑﺎ ﺍﻧﺪﺍﺯﺓ‬
‫ﭘﻠﻪ ﻫﺎﻱ ﻣﺨﺘﻠﻒ(‪ .‬ﺍﻧﺪﺍﺯﻩ ﭘﻠﻪ ﺑﺮﺍﻱ ﺧﻄﺎﻱ ﭘﻴﺶ ﺑﻴﻨﻲ ‪ DC‬ﻭ ﺩﻳﮕﺮ ﺿﺮﺍﻳﺐ ‪ AC‬ﺩﺭ ﻳﻚ ﻣﺎﺗﺮﻳﺲ ﻧﺮﻣﺎﻟﻴﺰﺍﺳﻴﻮﻥ ﻣﺸﺨﺺ ﻣﻲ ﺷﻮﺩ‪.‬‬
‫‪8‬‬
‫‪CE 342 – Multimedia HW# 5‬‬
‫‪H. Rabiee, Spring 2008‬‬
‫ﻣﺎﺗﺮﻳﺲ ﺑﻪ ﺧﺼﻮﺹ ﻣﻮﺭﺩ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺗﻮﺍﻧﺪ ﺩﺭ ﺍﺑﺘﺪﺍﻱ ﺟﺮﻳﺎﻥ ﺑﻴﺖ ﻓﺸﺮﺩﻩ ﺷﺪﻩ ﺑﻪ ﻋﻨﻮﺍﻥ ﺍﻃﻼﻋﺎﺕ ﺟﺎﻧﺒﻲ ﺑﻴﺎﻳﺪ‪ .‬ﺑﻪ ﻋﻨﻮﺍﻥ ﺟﺎﻳﮕﺰﻳﻦ‪ ،‬ﻫﺮ‬
‫ﻛﺲ ﻣﻲﺗﻮﺍﻧﺪ ﻣﺎﺗﺮﻳﺲ ﺗﻮﺻﻴﻪ ﺷﺪﻩ ﺗﻮﺳﻂ ‪ JPEG‬ﺭﺍ ﻛﻪ ﺩﺭ ﺷﻜﻞ ‪ ١١‬ﻧﺸﺎﻥ ﺩﺍﺩﻩ ﺷﺪﻩ ﺍﺳﺖ ﺑﻪ ﻃﻮﺭ ﭘﻴﺶ ﻓﺮﺽ ﺑﻪ ﻛﺎﺭ ﺑﺒﺮﺩ‪ .‬ﻣﻌﻤﻮﻻﹰ ﻣﻲ‬
‫ﺗﻮﺍﻥ ﺑﻴﻦ ﻛﻴﻔﻴﺖ ﻭ ﻧﺮﺥ ﺑﻴﺖ ﻣﺼﺎﻟﺤﻪ ﺑﺮ ﻗﺮﺍﺭ ﻛﺮﺩ )ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻓﺎﻛﺘﻮﺭ ﻣﻘﻴﺎﺱ(‪ .‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ ،‬ﻳﻚ ﻓﺎﻛﺘﻮﺭ ﻣﻘﻴﺎﺱ ﺑﺮﺍﺑﺮ ‪ ٢‬ﺑﻪ ﺍﻳﻦ‬
‫ﻣﻌﻨﺎﺳﺖ ﻛﻪ ﻃﻮﻝ ﭘﻠﻪ ﺑﺮﺍﻱ ﺗﻤﺎﻡ ﺿﺮﺍﻳﺐ ‪ ٢‬ﺑﺮﺍﺑﺮ ﺑﺸﻮﺩ‪ .‬ﺍﺯ ﺳﻮﻱ ﺩﻳﮕﺮ‪ ،‬ﻓﺎﻛﺘﻮﺭ ﻣﻘﻴﺎﺱ ‪٠/٥‬ﻛﻞ ﺍﻧﺪﺍﺯﻩ ﭘﻠﻪ ﻫﺎ ﺭﺍ ﻧﺼﻒ ﻣﻲﻛﻨﺪ‪ .‬ﻓﺎﻛﺘﻮﺭ‬
‫ﻣﻘﻴﺎﺱ ﻛﻮﭼﻜﺘﺮ‪ ،‬ﻧﻤﺎﻳﺶ ﺩﻗﻴﻘﺘﺮﻱ ﺍﺯ ﺗﺼﻮﻳﺮ ﺍﺻﻠﻲ ﺭﺍ ﻧﺘﻴﺠﻪ ﻣﻲ ﺩﻫﺪ ﻭﻟﻲ ﺑﻬﺮﺓ ﻓﺸﺮﺩﻩﺳﺎﺯﻱ ﺭﺍ ﻛﻢ ﻣﻲﻛﻨﺪ‪.‬‬
‫ﺑﺮﺍﻱ ﻛﺪ ﮔﺸﺎﻳﻲ ﺩﻭﺩﻭﻳﻲ ﺧﻄﺎﻱ ﭘﻴﺶ ﺑﻴﻨﻲ ‪ DC‬ﻛﻮﺍﻧﺘﻴﺰﻩ ﺷﺪﻩ ﻭ ﺿﺮﺍﻳﺐ ‪ ،DC‬ﺗﺮﻛﻴﺒﻲ ﺍﺯ ﻛﺪﻳﻨﮓ ﺑﺎ ﻃﻮﻝ ﻣﺘﻐﻴﺮ ﻭ ﻃﻮﻝ ﺛﺎﺑﺖ ﺑﻪ ﻛﺎﺭ‬
‫ﻣﻲﺭﻭﺩ‪ .‬ﺗﻤﺎﻡ ﻣﻘﺎﺩﻳﺮ ﺧﻄﺎﻱ ﭘﻴﺶ ﺑﻴﻨﻲ ﻣﻤﻜﻦ ﻭ ﺗﻤﺎﻡ ﻣﻘﺎﺩﻳﺮ ﺿﺮﺍﻳﺐ ‪ AC‬ﺑﻌﺪ ﺍﺯ ﻛﻮﺍﻧﺘﻴﺰﺍﺳﻴﻮﻥ ﺑﻪ ﺩﺳﺘﻪﻫﺎﻱ ﻣﺨﺘﻠﻒ )ﺑﺮ ﺍﺳﺎﺱ‬
‫ﺍﻧﺪﺍﺯﻩﺷﺎﻥ( ﺗﻘﺴﻴﻢ ﻣﻲ ﺷﻮﻧﺪ‪ .‬ﺟﺪﻭﻝ ‪ ٦-١٤‬ﺩﺭ ]‪ [١‬ﺭﺍ ﺑﺮﺍﻱ ﻗﺎﻧﻮﻥ ﺩﺳﺘﻪ ﺑﻨﺪﻱ ﺑﺒﻴﻨﻴﺪ‪ .‬ﺧﻄﺎﻱ ﭘﻴﺶ ﺑﻴﻨﻲ ‪ DC‬ﺑﺎ ﻳﻚ ﻛﺪ ﺩﻭ ﻗﺴﻤﺘﻲ ﻛﺪ‬
‫ﮔﺬﺍﺭﻱ ﻣﻲ ﺷﻮﺩ‪ .‬ﻗﺴﻤﺖ ﺍﻭﻝ ﻣﺸﺨﺺ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﺩﺭ ﻛﺪﺍﻡ ﺩﺳﺘﻪ ﺍﺳﺖ ﻭ ﻗﺴﻤﺖ ﺩﻭﻡ ﺍﻧﺪﺍﺯﻩ ﻧﺴﺒﻲ ﻭﺍﻗﻌﻲ ﺩﺭ ﺁﻥ ﺩﺳﺘﻪ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲﺩﻫﺪ‪.‬‬
‫ﻗﺴﻤﺖ ﺍﻭﻝ ﺑﺮ ﺍﺳﺎﺱ ﻓﺮﻛﺎﻧﺲ ﻫﺎﻱ ﺩﺳﺘﻪ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﻛﺪ ﻫﺎﻓﻤﻦ ﻣﻲ ﺷﻮﺩ ﻭ ﺩﺭ ﺣﺎﻟﻲ ﻛﻪ ﻗﺴﻤﺖ ﺩﻭﻡ ﺑﺎ ﻳﻚ ﻛﻠﻤﻪ ﻛﺪ ﺑﺎ ﻃﻮﻝ ﺛﺎﺑﺖ ﻛﺪ‬
‫ﻣﻲﺷﻮﺩ‪ .‬ﺑﺮﺍﻱ ﺿﺮﺍﻳﺐ ‪ ،AC‬ﻳﻚ ﺭﻭﺵ ﻛﺪﻳﻨﮓ ‪ RLC‬ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ‪ .‬ﺩﺭ ﺍﻳﻦ ﺭﻭﺵ‪ ،‬ﺍﺑﺘﺪﺍ ﺿﺮﺍﻳﺐ ‪ DCT‬ﺑﻪ ﻳﻚ ﺁﺭﺍﻳﻪ ﻳﻚ ﺑﻌﺪﻱ ﺑﺮ‬
‫ﺍﺳﺎﺱ ﺗﺮﺗﻴﺐ ﺯﻳﮕﺰﺍﮔﻲ ﻫﻤﺎﻥ ﻃﻮﺭ ﻛﻪ ﺩﺭ ﺷﻜﻞ ‪ ١٢‬ﻧﺸﺎﻥ ﺩﺍﺩﻩ ﺷﺪﻩ ﺍﺳﺖ ﺗﺒﺪﻳﻞ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬
‫ﺷﮑﻞ ‪ :11‬ﯾﮏ ﻣﺎﺗﺮﯾﺲ ﻧﺮﻣﺎل ﺳﺎزي ﻣﻌﻤﻮل‬
‫ﺷﮑﻞ‪ :12‬ﯾﮏ ﻣﺎﺳﮏ زﯾﮕﺰاگ ﻣﻌﻤﻮل‬
‫ﺿﺮﺍﻳﺐ ‪ AC‬ﺑﻪ ﺳﻤﺒﻮﻝ ﻫﺎ ﺗﺒﺪﻳﻞ ﻣﻲ ﺷﻮﺩ ﻭ ﻫﺮ ﺳﻤﺒﻮﻝ ﻳﻚ ﺟﻔﺖ ﺗﺸﻜﻴﻞ ﻳﺎﻓﺘﻪ ﺍﺯ ﺍﻧﺪﺍﺯﺓ ﺻﻔﺮﻫﺎ ﺍﺯ ﺁﺧﺮﻳﻦ ﻣﻘﺪﺍﺭ ﻏﻴﺮ ﺻﻔﺮ ﺗﺎ ﻣﻘﺪﺍﺭ‬
‫ﻏﻴﺮ ﺻﻔﺮ ﺑﻌﺪﻱ ﺍﺳﺖ‪ .‬ﻣﻘﺪﺍﺭ ﻏﻴﺮ ﺻﻔﺮ ﺗﻮﺳﻂ ﺩﻭ ﻗﺴﻤﺖ ﻣﺸﺨﺺ ﻣﻲ ﺷﻮﺩ‪ ،‬ﺷﺒﻴﻪ ﻛﺎﺭﻱ ﻛﻪ ﺑﺮﺍﻱ ﺧﻄﺎﻱ ﭘﻴﺶ ﺑﻴﻨﻲ ‪ DC‬ﺍﻧﺠﺎﻡ ﺷﺪ‪.‬‬
‫ﻗﺴﻤﺖ ﺍﻭﻝ ﻣﺸﺨﺺ ﻣﻲﻛﻨﺪ ﻛﻪ ﺑﻪ ﻛﺪﺍﻡ ﺩﺳﺘﻪ ﺗﻌﻠﻖ ﺩﺍﺭﺩ ﻭ ﻗﺴﻤﺖ ﺩﻭﻡ ﺍﻧﺪﺍﺯﻩ ﻧﺴﺒﻲ ﺭﺍ ﻣﺸﺨﺺ ﻣﻲﻛﻨﺪ‪ .‬ﺳﻤﺒﻮﻝ ﺗﺸﻜﻴﻞ ﺷﺪﻩ ﺍﺯ ‪RLC‬‬
‫ﺻﻔﺮ ﻭ ﺩﺳﺘﻪ ﻣﻘﺪﺍﺭ ﻏﻴﺮ ﺻﻔﺮ‪ ،‬ﻛﺪ ﻫﺎﻓﻤﻦ ﻣﻲ ﺷﻮﺩ‪ ،‬ﺩﺭ ﺣﺎﻟﻲ ﻛﻪ ﺍﻧﺪﺍﺯﻩ ﻧﺴﺒﻲ ﻣﻘﺪﺍﺭ ﻏﻴﺮ ﺻﻔﺮ ﺑﺎ ﻛﺪ ﺑﺎ ﻃﻮﻝ ﺛﺎﺑﺖ ﻛﺪ ﻣﻲ ﺷﻮﺩ‪ .‬ﺍﺳﺘﺎﻧﺪﺍﺭﺩ‪،‬‬
‫ﺑﻌﻀﻲ ﺟﺪﺍﻭﻝ ﻫﺎﻓﻤﻦ ﭘﻴﺶ ﻓﺮﺽ ﺭﺍ ﺑﺮﺍﻱ ﺧﻄﺎﻱ ﭘﻴﺶﺑﻴﻨﻲ ‪ DC‬ﻭ ﺳﻤﺒﻮﻝﻫﺎﻱ ‪ AC‬ﺗﻮﺻﻴﻪ ﻣﻲ ﻛﻨﺪ‪ .‬ﻭﻟﻲ ﻛﺎﺭﺑﺮ ﻣﻲﺗﻮﺍﻧﺪ ﺟﺪﺍﻭﻝ‬
‫ﻣﺨﺘﻠﻔﻲ ﻛﻪ ﺑﺮﺍﻱ ﺧﺼﻮﺻﻴﺎﺕ ﺁﻣﺎﺭﻱ ﺗﺼﺎﻭﻳﺮ ﺩﺭ ﻳﻚ ﻛﺎﺭﺑﺮﺩ ﺧﺎﺹ ﺑﻬﻴﻨﻪ ﺷﺪﻩ ﺍﻧﺪ ﺭﺍ ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﺪ‪ .‬ﺑﺮﺍﻱ ﻳﻚ ﺗﺼﻮﻳﺮ ﺭﻧﮕﻲ‪ ،‬ﻫﺮ ﻣﻮﻟﻔﺔ‬
‫ﺭﻧﮓ ﺑﻪ ﻃﻮﺭ ﺟﺪﺍﮔﺎﻧﻪ ﺑﺎ ﻫﻤﻴﻦ ﺭﻭﺵ ﻓﺸﺮﺩﻩ ﻣﻲﺷﻮﺩ‪ .‬ﺷﻜﻞ ‪ ١٣‬ﺗﺼﺎﻭﻳﺮ ﻓﺸﺮﺩﻩ ﺷﺪﻩ ‪ JPEG‬ﺑﻪ ﺩﺳﺖ ﺁﻣﺪﻩ ﺑﺎ ﺍﻧﺘﺨﺎﺏ ﻓﺎﻛﺘﻮﺭﻫﺎﻱ ﻣﻘﻴﺎﺱ‬
‫ﻣﺨﺘﻠﻒ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ‪.‬‬
‫‪9‬‬
‫‪CE 342 – Multimedia HW# 5‬‬
‫‪H. Rabiee, Spring 2008‬‬
‫ﺷﮑﻞ ‪ : 13‬ﺗﺼﺎوﯾﺮ ﻓﺸﺮده ﺷﺪه ‪ JPEG‬ﺑﻪ دﺳﺖ آﻣﺪه ﺑﺎ اﻧﺘﺨﺎب ﻓﺎﮐﺘﻮرﻫﺎي ﻣﻘﯿﺎس ﻣﺨﺘﻠﻒ‬
‫‪ -4-2‬ﮐﻮاﻧﺘﯿﺰاﺳﯿﻮن ﺑﺮداري‬
‫ﻛﻮﺍﻧﺘﻴﺰﺍﺳﻴﻮﻥ ﺑﺮﺩﺍﺭﻱ )‪ (VQ‬ﺭﺍﻫﺒﺮﺩ ﺩﻳﮕﺮﻱ ﺑﺮﺍﻱ ﻓﺸﺮﺩﻩ ﺳﺎﺯﻱ ﺗﺼﻮﻳﺮ ﺍﺳﺖ‪ .‬ﺍﻳﺪﻩ ﭘﺸﺖ ‪ VQ‬ﺗﻌﻴﻴﻦ ﺑﻬﺘﺮﻳﻦ ﺩﺳﺘﻪ ﺍﺯ ﺍﻟﮕﻮﻫﺎﻱ ﺑﻠﻮﻙ‬
‫ﭘﺎﻳﻪ )ﻛﻪ ﻫﺮ ﻛﺪﺍﻡ ﻳﻚ ﻛﻠﻤﻪ ﻛﺪ ﺧﻮﺍﻧﺪﻩ ﻣﻲ ﺷﻮﻧﺪ( ﺍﺳﺖ ﻛﻪ ﺑﻪ ﺑﻬﺘﺮﻳﻦ ﻭﺟﻪ ﺗﻤﺎﻡ ﺑﻠﻮﻙ ﻫﺎﻱ ﺗﺼﺎﻭﻳﺮ ﺭﺍ ﺩﺭ ﻳﻚ ﺗﺼﻮﻳﺮ ﻧﻤﺎﻳﺶ ﺩﻫﺪ‪.‬‬
‫ﻣﺠﻤﻮﻋﻪ ﺗﻤﺎﻡ ﺍﻟﮕﻮﻫﺎﻱ ﭘﺎﻳﻪ‪ ،‬ﻛﺘﺎﺏ – ﻛﺪ ﺧﻮﺍﻧﺪﻩ ﻣﻲ ﺷﻮﺩ‪ .‬ﻣﻌﻤﻮﻻﹰ ﺑﻬﺘﺮﻳﻦ ﻛﺘﺎﺏ – ﻛﺪ ﺑﺮﺍﻱ ﻳﻚ ﻛﻼﺱ ﺍﺯ ﺗﺼﺎﻭﻳﺮ‪ ،‬ﺍﺯ ﻗﺒﻞ ﻃﺮﺍﺣﻲ ﻣﻲ‬
‫ﺷﻮﺩ‪ .‬ﺩﺭ ﻓﺮﺍﻳﻨﺪ ﻛﺪﮔﺬﺍﺭﻱ ﺑﺮﺍﻱ ﻫﺮ ﺑﻠﻮﻙ ﺗﺼﻮﻳﺮ ﺩﺍﺩﻩ ﺷﺪﻩ ﺑﻬﺘﺮﻳﻦ ﻛﻠﻤﻪ – ﻛﺪ ﻣﻨﻄﺒﻖ ﺩﺭ ﻛﺘﺎﺏ‪ -‬ﻛﺪ ﭘﻴﺪﺍ ﻣﻲ ﺷﻮﺩ‪ .‬ﺍﮔﺮ ﭼﻪ ﺍﻳﻦ ﺭﻭﺵ‬
‫ﺟﺴﺘﺠﻮ ﻣﺤﺎﺳﺒﺎﺕ ﺑﻴﺸﺘﺮﻱ ﺭﺍ ﻧﺴﺒﺖ ﺑﻪ ‪ JPEG‬ﻃﻠﺐ ﻣﻲﻛﻨﺪ‪ ،‬ﺩﺭ ﻛﺪﮔﺸﺎﻳﻲ ﺑﺴﻴﺎﺭ ﺳﺮﻳﻌﺘﺮ ﻭ ﺳﺎﺩﻩﺗﺮ ﺍﺳﺖ‪ .‬ﻛﺪﮔﺸﺎﻳﻲ ﺍﻃﻼﻋﺎﺕ‬
‫ﻛﺪﮔﺬﺍﺭﻱ ﺷﺪﻩ ﺑﺎ ﻛﻮﺍﻧﺘﻴﺰﺍﺳﻴﻮﻥ ﺑﺮﺩﺍﺭﻱ ﺷﺎﻣﻞ ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﻛﻠﻤﻪ – ﻛﺪ ﻣﻨﺎﺳﺐ ﺩﺭ ﻛﺘﺎﺏ – ﻛﺪ – ﺑﺮ ﺍﺳﺎﺱ ﺍﻧﺪﻳﺲ ﺍﻳﺠﺎﺩ ﺷﺪﻩ ﺩﺭ ﺧﻼﻝ‬
‫ﭘﺮﺩﺍﺯﺵ ﻛﺪﮔﺬﺍﺭﻱ ﺍﺳﺖ‪ .‬ﭘﻴﭽﻴﺪﮔﻲ ‪ ،VQ‬ﺑﻪ ﻃﻮﺭ ﻧﻤﺎﻳﻲ ﺑﺎ ﺍﻧﺪﺍﺯﻩ ﺑﻠﻮﻙ )ﺗﻌﺪﺍﺩ ﭘﻴﻜﺴﻞﻫﺎﻱ ﻫﺮ ﺑﻠﻮﻙ( ﺍﻓﺰﺍﻳﺶ ﻣﻲ ﻳﺎﺑﺪ‪ .‬ﺩﺭ ﻋﻤﻞ ﺍﻧﺪﺍﺯﻩ‬
‫ﺑﻠﻮﻙ ﻣﺴﺎﻭﻱ ﻳﺎ ﻛﻮﭼﻜﺘﺮ ﺍﺯ ‪ ٤*٤‬ﺍﺳﺖ ﻛﻪ ﺍﻏﻠﺐ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲﺷﻮﺩ‪ .‬ﺷﻜﻞ ‪ ،١٤‬ﻣﻌﻤﺎﺭﻱ ﻛﻮﺍﻧﺘﻴﺰﺍﺳﻴﻮﻥ ﺑﺮﺩﺍﺭﻱ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲﺩﻫﺪ‪.‬‬
‫ﺷﮑﻞ ‪ :14‬ﮐﻮاﻧﺘﯿﺰاﺳﯿﻮن ﺑﺮداري‬
‫ﺳﺎﺩﻩﺗﺮﻳﻦ ﺭﺍﻫﺒﺮﺩ ﺑﺮﺍﻱ ﭘﺮﺩﺍﺯﺵ ﺑﺨﺶﻫﺎﻱ ﺗﺼﻮﻳﺮ‪ ،‬ﺗﻘﺴﻴﻢ ﺗﺼﻮﻳﺮ ﻭﺭﻭﺩﻱ ﺩﺭ ﻛﺪﮔﺬﺍﺭ ﺑﻪ ﺑﻠﻮﻙﻫﺎﻱ ﻳﺎ ﺑﺮﺩﺍﺭﻫﺎﻱ ﻣﺴﺘﻄﻴﻠﻲ‪ ،‬ﻏﻴﺮ‬
‫ﻫﻤﭙﻮﺷﺎﻧﻲ‪ ،‬ﭼﺴﺒﻴﺪﻩ ﻭ ﻛﻮﭼﻚ ﺍﺯ ﭘﻴﻜﺴﻞﻫﺎ ﺍﺳﺖ ﻛﻪ ﻫﺮ ﻛﺪﺍﻡ ﺟﺪﺍﮔﺎﻧﻪ ﻛﻮﺍﻧﺘﻴﺰﻩ ﺷﻮﺩ‪ .‬ﺍﺑﻌﺎﺩ ﺑﺮﺩﺍﺭ ﻣﺴﺎﻭﻱ ﺗﻌﺪﺍﺩ ﭘﻴﻜﺴﻞ ﻫﺎﻱ ﻫﺮ ﺑﻠﻮﻙ‬
‫‪10‬‬
‫‪CE 342 – Multimedia HW# 5‬‬
‫‪H. Rabiee, Spring 2008‬‬
‫ﺍﺳﺖ‪ .‬ﺑﺮﺩﺍﺭ ﻧﻤﻮﻧﻪ ﻫﺎ‪ ،‬ﺍﻟﮕﻮﻳﻲ ﺍﺳﺖ ﻛﻪ ﺑﺎﻳﺪ ﺑﺎ ﻣﺠﻤﻮﻋﻪ ﺍﻟﮕﻮﻫﺎﻱ ﺗﺴﺖ ﺗﻘﺮﻳﺐ ﺯﺩﻩ ﺷﻮﺩ‪ .‬ﺍﻟﮕﻮﻫﺎ ﺩﺭ ﻳﻚ ﺩﻳﻜﺸﻨﺮﻱ ﺫﺧﻴﺮﻩ ﻣﻲﺷﻮﻧﺪ‪.‬‬
‫ﺍﻟﮕﻮﻫﺎﻱ ﻣﻮﺟﻮﺩ ﺩﺭ ﻛﺘﺎﺏ – ﻛﺪ‪ ،‬ﻛﻠﻤﻪ – ﻛﺪ ﺧﻮﺍﻧﺪﻩ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬
‫ﺩﺭ ﺧﻼﻝ ﻓﺸﺮﺩﻩﺳﺎﺯﻱ‪ ،‬ﻛﺪﮔﺬﺍﺭ‪ ،‬ﺑﻪ ﻫﺮ ﺩﻭ ﺑﺮﺩﺍﺭ ﻭﺭﻭﺩﻱ ﻳﻚ ﺁﺩﺭﺱ ﻳﺎ ﺍﻧﺪﻳﺲ ﻣﻲ ﺩﻫﺪ ﻛﻪ ﻛﻠﻤﻪ – ﻛﺪﻱ ﺭﺍ ﻛﻪ ﺩﺭ ﻛﺘﺎﺏ – ﻛﺪ ﺑﻪ ﺑﻬﺘﺮﻳﻦ‬
‫ﻭﺟﻪ ﺑﺮﺩﺍﺭ ﻭﺭﻭﺩﻱ ﺭﺍ ﺗﻘﺮﻳﺐ ﻣﻲﺯﻧﺪ ﻣﺸﺨﺺ ﻣﻲﻛﻨﺪ‪ .‬ﺩﺭ ﻛﺪﮔﺬﺍﺭ ﻫﺮ ﺍﻧﺪﻳﺲ ﺑﻪ ﻳﻚ ﺑﺮﺩﺍﺭ ﺧﺮﻭﺟﻲ ﮔﺮﻓﺘﻪ ﺷﺪﻩ ﺍﺯ ﻛﺘﺎﺏ – ﻛﺪ ﻣﺘﻨﺎﻇﺮ‬
‫ﻣﻲﺷﻮﺩ‪ .‬ﻛﺪﮔﺸﺎ ﺗﺼﻮﻳﺮ ﺭﺍ ﺑﺎ ﻫﻤﺎﻥ ﺗﺮﺗﻴﺐ ﺑﺮﺩﺍﺭﻫﺎﻱ ﻭﺭﻭﺩﻱ ﺑﺎﺯﺳﺎﺯﻱ ﻣﻲﻛﻨﺪ‪ .‬ﻭﻳﺮﺍﻳﺶ ﺳﻮﻡ ‪ Indeo‬ﺍﺯ ﺷﺮﮐﺖ ‪ ،Intel‬ﺑﺮ ﺍﺳﺎﺱ ﺭﻭﺵ‬
‫ﻛﻮﺍﻧﺘﻴﺰﺍﺳﻴﻮﻥ ﺑﺮﺩﺍﺭﻱ ﻛﺎﺭ ﻣﻲ ﻛﻨﺪ‪.‬‬
‫‪ -3‬آزﻣﺎﯾﺶ‬
‫‪ -١‬ﺍﺯ ﻳﻚ ﺗﺼﻮﻳﺮ ‪ bmp‬ﺑﻪ ﻋﻨﻮﺍﻥ ﺗﺼﻮﻳﺮ ﻭﺭﻭﺩﻱ ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﻴﺪ ﻭ ﻛﺪﻳﻨﮓ ‪ RLC‬ﺩﻭ ﺳﻄﺤﻲ ﺭﺍ ﺑﻪ ﻃﻮﺭ ﺩﺳﺘﻲ ﺍﻧﺠﺎﻡ ﺩﻫﻴﺪ ﻭ ﺍﺣﺘﻤﺎﻝ‬
‫ﻃﻮﻝ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺭﺍ ﺣﺴﺎﺏ ﻛﻨﻴﺪ‪ .‬ﺳﭙﺲ ﻛﺪﻳﻨﮓ ﻫﺎﻓﻤﻦ ﺭﺍ ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﻴﺪ )ﺑﻪ ﺷﻜﻞ ‪ ٣‬ﻣﺮﺍﺟﻌﻪ ﻛﻨﻴﺪ( ﺗﺎ ﻛﻠﻤﻪ ﻛﺪ ﻫﺮ ﺳﻤﺒﻮﻝ ﺭﺍ ﺑﻪ ﺩﺳﺖ‬
‫ﺁﻭﺭﻳﺪ‪ .‬ﺳﭙﺲ ﻃﻮﻝ ﻣﻴﺎﻧﮕﻴﻦ ﻛﺪ ﺭﺍ ﻣﺤﺎﺳﺒﻪ ﻛﻨﻴﺪ‪ .‬ﻣﻲﺗﻮﺍﻧﻴﺪ ﺍﺯ ﺩﺳﺘﻮﺭ )(‪ bmpread‬ﺑﺮﺍﻱ ﺧﻮﺍﻧﺪﻥ ﻭ ﻧﻤﺎﻳﺶ ﻓﺎﻳﻞ ﻭﺭﻭﺩﻱ ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﻴﺪ‪.‬‬
‫‪ -٢‬ﺑﺎ ‪ dctdemo.m‬ﺑﺎﺯﻱ ﻛﻨﻴﺪ‪ .‬ﺑﻪ ﺗﺄﺛﻴﺮ ﺍﻧﺘﺨﺎﺏ ﺗﻌﺪﺍﺩ ﻣﺨﺘﻠﻒ ﺿﺮﺍﻳﺐ ‪ DCT‬ﺗﻮﺟﻪ ﻛﻨﻴﺪ‪ .‬ﺑﺮﺍﻱ ﻳﻚ ﺗﺼﻮﻳﺮ ﺍﻧﺘﺨﺎﺏ ﺷﺪﻩ‪ ،‬ﺣﺪﺍﻗﻞ‬
‫ﺗﻌﺪﺍﺩ ﺿﺮﺍﻳﺒﻲ ﻛﻪ ﺑﺎﻳﺪ ﺍﻧﺘﺨﺎﺏ ﻛﻨﻴﺪ ﺗﺎ ﻧﺘﻴﺠﻪ ﺭﺿﺎﻳﺖ ﺑﺨﺶ ﺑﮕﻴﺮﻳﺪ ﭼﻴﺴﺖ؟ ﺑﺮﺍﻱ ﺍﻳﻦ ﻣﻨﻈﻮﺭ ﺑﻪ ﺗﺼﻮﻳﺮ ﺧﻄﺎ ﻧﻴﺰ ﺗﻮﺟﻪ ﻛﻨﻴﺪ‪.‬‬
‫‪ -٣‬ﭘﻴﻮﺳﺖ ﺍﻟﻒ ﺑﺮﻧﺎﻣﺔ ‪ dctquant‬ﺭﺍ ﻣﻲ ﺩﻫﺪ ﻛﻪ ‪ DCT‬ﺭﺍ ﺑﺮ ﺭﻭﻱ ﺑﻠﻮﻙ ‪ ٨*٨‬ﺣﺴﺎﺏ ﻣﻲ ﻛﻨﺪ ﻭ ﺳﭙﺲ ﺿﺮﺍﻳﺐ ‪ DCT‬ﺭﺍ ﻛﻮﺍﻧﺘﻴﺰﻩ‬
‫ﻣﻲ ﻛﻨﺪ‪ .‬ﺳﭙﺲ ﻋﻜﺲ ‪ DCT‬ﺭﺍ ﺍﻧﺠﺎﻡ ﻣﻲ ﺩﻫﺪ ﺗﺎ ﺗﺼﻮﻳﺮ ﺑﺎﺯﺳﺎﺯﻱ ﺷﺪﻩ ﺭﺍ ﺑﻪ ﺩﺳﺖ ﺁﻭﺭﺩ‪ .‬ﺑﺎ ﺑﺮﻧﺎﻣﻪ ﺑﺎﺯﻱ ﻛﻨﻴﺪ ﺗﺎ ﺗﺄﺛﻴﺮ ﻛﻮﺍﻧﺘﻴﺰﺍﺳﻴﻮﻥ ﺭﺍ ﺑﺎ‬
‫ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻓﺎﻛﺘﻮﺭﻫﺎﻱ ﻣﻘﻴﺎﺱ ﻣﺨﺘﻠﻒ ﺑﻪ ﺩﺳﺖ ﺁﻭﺭﻳﺪ‪ .‬ﺑﺮﺍﻱ ﻓﻬﻤﻴﺪﻥ ﺍﻳﻦ ﺑﺮﻧﺎﻣﻪ ﺷﻤﺎ ﺑﺎﻳﺪ ﺗﻮﺍﺑﻊ ”‪ “blkproc‬ﻭ ”‪ “mask2.m‬ﺭﺍ ﻳﺎﺩ‬
‫ﺑﮕﻴﺮﻳﺪ‪.‬‬
‫‪ dctquant.m -٤‬ﺭﺍ ﺍﺻﻼﺡ ﻛﻨﻴﺪ ﻭ ﺑﻪ ﺟﺎﻱ ﻛﻮﺍﻧﺘﻴﺰﻩ ﻛﺮﺩﻥ ﺿﺮﺍﻳﺐ ‪ DCT‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻣﺎﺗﺮﻳﺲ ﻧﺮﻣﺎﻟﻴﺰﺍﺳﻴﻮﻥ ﻓﺮﺍﻫﻢ ﺷﺪﻩ ﺗﻨﻬﺎ ‪L‬‬
‫ﺿﺮﻳﺐ ﺍﻭﻝ ﺭﺍ ﺩﺭ ﻳﻚ ﺗﺮﺗﻴﺐ ﺯﻳﮓ ﺯﺍﮒ ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﻴﺪ‪ .‬ﻣﻘﺎﺩﻳﺮ ﻣﺨﺘﻠﻒ ‪ L‬ﺑﻴﻦ ‪ ١‬ﺗﺎ ‪ ١٦‬ﺭﺍ ﺍﻣﺘﺤﺎﻥ ﻛﻨﻴﺪ‪.‬‬
‫ﺗﻮﺟﻪ ‪ :‬ﺷﻤﺎ ﺑﺎﻳﺪ ﻣﺎﺳﻚ ﺭﺍ ﺩﺭ ﺑﺮﻧﺎﻣﻪ ﻃﻮﺭﻱ ﺍﺻﻼﺡ ﻛﻨﻴﺪ ﻛﻪ ﺑﺘﻮﺍﻧﺪ ﺗﺸﺨﻴﺺ ﺩﻫﺪ ﻛﻪ ﻛﺪﺍﻡ ﺿﺮﺍﻳﺐ ﺑﺎﻳﺪ ﺣﻔﻆ ﺷﻮﻧﺪ ﻭ ﻛﺪﺍﻡ ﺿﺮﺍﻳﺐ ﺑﺎﻳﺪ‬
‫ﺻﻔﺮ ﺷﻮﻧﺪ‪ .‬ﻳﺎ ﺍﻳﻨﻜﻪ ﻣﻲ ﺗﻮﺍﻧﻴﺪ ﻳﻚ ﺑﺮﻧﺎﻣﻪ ”‪ “mask‬ﺟﺪﻳﺪ ﺑﺮﺍﻱ ﺍﻳﻦ ﻛﺎﺭ ﺑﻨﻮﻳﺴﻴﺪ‪.‬‬
‫‪ -٥‬ﺑﺮﻧﺎﻣﻪ ‪ dctquant.m‬ﺭﺍ ﺍﺻﻼﺡ ﻛﻨﻴﺪ ﻃﻮﺭﻱ ﻛﻪ ﺑﻪ ﺟﺎﻱ ﻛﻮﺍﻧﺘﻴﺰﻩ ﻛﺮﺩﻥ ﺿﺮﺍﻳﺐ ‪ DCT‬ﺑﺮ ﺍﺳﺎﺱ ﻣﺎﺗﺮﻳﺲ ﻧﺮﻣﺎﻟﻴﺰﺍﺳﻴﻮﻥ‪ ،‬ﺗﻨﻬﺎ ﺿﺮﺍﻳﺒﻲ‬
‫ﺭﺍ ﻧﮕﻪ ﺩﺍﺭﺩ ﻛﻪ ﺍﻧﺪﺍﺯﻩ ﺍﻱ ﺑﺰﺭﮔﺘﺮ ﺍﺯ ﺁﺳﺘﺎﻧﻪ ‪ T‬ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ‪ .‬ﻣﻘﺎﺩﻳﺮ ﻣﺨﺘﻠﻒ ‪ T‬ﺑﻴﻦ ‪ ١‬ﺗﺎ ‪ ٢٥٦‬ﺭﺍ ﺍﻣﺘﺤﺎﻥ ﻛﻨﻴﺪ‪ .‬ﻧﺘﻴﺠﻪﻱ ﺣﺎﺻﻞ ﺍﺯ ﺍﻳﻦ‬
‫ﺁﺯﻣﺎﻳﺶ ﺭﺍ ﺑﺎ ﻧﺘﻴﺠﻪ ﺁﺯﻣﺎﻳﺶ ﺷﻤﺎﺭﻩ ‪ ٤‬ﻣﻘﺎﻳﺴﻪ ﻛﻨﻴﺪ‪.‬‬
‫‪ -4‬ﮔﺰارش‬
‫‪ -١‬ﺩﻳﺎﮔﺮﺍﻡ ﻳﻚ ﺍﻟﮕﻮﺭﻳﺘﻢ ﻧﻤﻮﻧﻪ ﻛﺪﻳﻨﮓ ﺗﺼﻮﻳﺮ ﺭﺍ ﺑﻜﺸﻴﺪ )ﺑﻪ ﺷﻜﻞ ‪ ١‬ﻣﺮﺍﺟﻌﻪ ﻛﻨﻴﺪ( ﻭ ﺗﻮﺿﻴﺢ ﺩﻫﻴﺪ ﻛﺪﺍﻡ ﻗﺴﻤﺖ ﻫﺎ ﺑﺎ ﺗﻠﻔﺎﺕ ﻭ ﻛﺪﺍﻡ‬
‫ﺑﺪﻭﻥ ﺗﻠﻔﺎﺕ ﻫﺴﺘﻨﺪ‪.‬‬
‫‪ -٢‬ﺑﺮﻧﺎﻣﻪ ﻫﺎﻳﻲ ﻛﻪ ﻧﻮﺷﺘﻪ ﺍﻳﺪ ﺗﺤﻮﻳﻞ ﺑﺪﻫﻴﺪ‪ .‬ﻫﻢ ﻛﺪ ﺑﺮﻧﺎﻣﻪ ﻫﻢ ﺧﺮﻭﺟﻲ ﺁﻥ‪.‬‬
‫‪ -٣‬ﺑﺮﺍﻱ ﺁﺯﻣﺎﻳﺶ ﻫﺎﻱ ‪ ٢،٣،٤‬ﺍﺛﺮ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻣﺨﺘﻠﻒ ﻛﻮﺍﻧﺘﻴﺰﺍﺳﻴﻮﻥ ﺭﺍ ﺗﻮﺿﻴﺢ ﺩﻫﻴﺪ‪ ،‬ﺗﻌﺪﺍﺩ ﺿﺮﺍﻳﺐ ﺑﺎﻗﻴﻤﺎﻧﺪﻩ ﻭ ﺁﺳﺘﺎﻧﻪ ﺭﺍ ﺑﻨﻮﻳﺴﻴﺪ‪.‬‬
‫ﺣﺪﺍﻛﺜﺮ ﻓﺎﻛﺘﻮﺭ ﻣﻘﻴﺎﺱ‪ ،‬ﺣﺪﺍﻗﻞ ﺗﻌﺪﺍﺩ ﺿﺮﺍﻳﺐ‪ ،‬ﺣﺪﺍﻛﺜﺮ ﺁﺳﺘﺎﻧﻪ ﺭﺍ ﺑﻪ ﻃﻮﺭﻱ ﻛﻪ ﻛﻴﻔﻴﺖ ﺗﺼﻮﻳﺮ ﺭﺍ ﺩﺭ ﺣﺪ ﺭﺿﺎﻳﺖ ﺑﺨﺶ ﺣﻔﻆ ﻛﻨﻴﻢ‬
‫ﺑﻨﻮﻳﺴﻴﺪ‪.‬‬
‫‪ -٤‬ﺑﺮﺍﻱ ﻳﻚ ﺩﻧﺒﺎﻟﻪ ﻭﻳﺪﻳﻮ‪ ،‬ﺍﻧﺪﺍﺯﻩ ﺗﺼﻮﻳﺮ ﺩﺭ ﻫﺮ ﻗﺎﺏ ‪ ٣٦٠*٢٤٠‬ﻭ ﻧﺮﺥ ﻗﺎﺑﻞ ‪ ٣٠‬ﻗﺎﺏ ﺑﺮ ﺛﺎﻧﻴﻪ ﺑﺎ ﺭﻧﮓ ﻛﺎﻣﻞ )‪ ٣‬ﺑﺎﻳﺖ ﺑﺮ ﭘﻴﻜﺴﻞ( ﺩﺍﺭﺩ‪.‬‬
‫ﻧﺴﺒﺖ ﻓﺸﺮﺩﻩ ﺳﺎﺯﻱ ﺑﺮﺍﻱ ﺑﻪ ﺩﺳﺖ ﺁﻭﺭﺩﻥ ﺩﻧﺒﺎﻟﻪ ‪ ١/٥ mbits/second‬ﭼﻴﺴﺖ؟‬
‫‪ -5‬ﻣﺮاﺟﻊ‬
‫‪[1]. R. C. Gonzalez & R. E. Woods, “Digital Image Processing”, Addison Wesley, 1992.‬‬
‫‪11‬‬
‫‪CE 342 – Multimedia HW# 5‬‬
‫‪H. Rabiee, Spring 2008‬‬
[2]. A. N. Netravali and B. G. Haskell, “Digital Pictures, Representation, Compression, and
Standards”, 2nd ed., Plenum Press, 1995.
‫ﭘﯿﻮﺳﺖ اﻟﻒ‬
******************************************************************
* MATLAB Script file for demonstration of DCT representation of images *
* You should find out how to use “blkproc” by on-line help in matlab.
******************************************************************
function dctquant(FileName,dx,dy);
% usage : dctquant('h:\el593\exp10\lena.img',256,256); (WYT: please verify)
% Note, dctquant calls subfunctions mask2()
Img=fread(fopen(FileName),[dx,dy]);
colormap(gray(256));
image(Img');
set(gca,'XTick',[],'YTick',[]);
title('Original Image');
truesize;
drawnow
y=blkproc(Img,[8 8],'dct2');
yy=blkproc(y,[8 8],'mask2');
yq=blkproc(yy,[8,8],'idct2');
figure;
colormap(gray(256));
image(yq');
set(gca,'XTick',[],'YTick',[]);
title('Quantized Image');
truesize;
drawnow;
CE 342 – Multimedia HW# 5
H. Rabiee, Spring 2008
12
‫ﭘﯿﻮﺳﺖ ب‬
******************************************************************
* MATLAB Script file for demonstration of DCT (subroutine 1) *
******************************************************************
function [y]=mask2(x);
mask=[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 22 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 56 112 100 103 99];
% Normally c=1
c=16;
mask=c*mask;
z=round(x./mask);
y=mask.*z;
CE 342 – Multimedia HW# 5
H. Rabiee, Spring 2008
13