Yırtılabilir zar modeli yardımıyla görüntü sıkıştırma

1995
Gençata, Ayşegül
Yayınevi
Fen Bilimleri Enstitüsü
Özet
The digital representation of an image requires a very large number of bits. Pictorial data is produced by sampling in space and quantizing in brightness analog scenes. The sampling step size is usually chosen small enough to avoid interpolation before display and relies on the integration ability of the human visual system. A digital image is an N by N array of integer numbers ör picture elements (pixels). The goal of image coding is to reduce (to compress), as much as possible, the number of bits necessary to represent and reconstruct a faithful duplicate of the original picture. Any data originated from an image, are not random. Adjacent pixels have similar gray values,this is an important spatial correlation. If this correlation is exploited in an appropriate way, the number of N2B bits required for represen¬ tation of a digital image can be reduced. The set of techniques based on this classical view of image coding problem is called "first generation" [1]. By using information theory various alternatives were proposed, these coding techniques reached a compression ratio at about 10 to 1. This certainly does not mean that the upper bound given by the entropy of the source has also been reached. First, the entropy of an image is unknown and depends heavily on the model used, and second, information theory is not customarily used to take into account what the human eye sees and how it sees. The very end of almost every image processing system is the human eye. Although our visual system is by far the best image processing system öne can think of, it is also far from being perfect. If the coding scheme is matched to the human visual system and tries to imitate its functions, high compressions can be expected. A few methods designed with these approaches give encouraging >+l) _ -(") " dE T duitj is the relaxation parameter and w G [0, 2]. n denotes the iteration number in increasing order. By taking derivatives and rearranging we obtain, u\n7L> - An) - A2 (141 [4wt-,j - (ui-ij + ui+ltj + Uij-i. + Uitj+i)} where /3ij = 0, and *4j = ^»'.i where fcj = 1. During the SOR iterations, pre-existing data should not be modified and iterative calculations must be performed only for non-existing data. If goal is to construct surface from sparse data, it is not important whether the data is at an edge location or in the middle of a smooth region. The important thing is to locate and use available data for reconstruction. The molecule of a membrane has a structure with five elements. This molecule has its center and the four connected components which are only one pixel apart. This means that any molecule cannot contain pixels in different regions separated by edges, each pixel in membrane functional is enough for reconstruction. The method is applied to various synthetic and real images. The compression ratio for syntetic images is approximately 40:1 and for real images approximately 20:1. Xlll ABSTRACT IMAGE COMPRESSION BY USING WEAK MEMBRANE MODEL The digital representation of an image requires a very large number of bits. Pictorial data is produced by sampling in space and quantizing in brightness analog scenes. The sampling step size is usually chosen small enough to avoid interpolation before display and relies on the integration ability of the human visual system. A digital image is an N by N array of integer numbers ör picture elements (pixels). The goal of image coding is to reduce (to compress), as much as possible, the number of bits necessary to represent and reconstruct a faithful duplicate of the original picture. Any data originated from an image, are not random. Adjacent pixels have similar gray values,this is an important spatial correlation. If this correlation is exploited in an appropriate way, the number of N2B bits required for represen¬ tation of a digital image can be reduced. The set of techniques based on this classical view of image coding problem is called "first generation" [1]. By using information theory various alternatives were proposed, these coding techniques reached a compression ratio at about 10 to 1. This certainly does not mean that the upper bound given by the entropy of the source has also been reached. First, the entropy of an image is unknown and depends heavily on the model used, and second, information theory is not customarily used to take into account what the human eye sees and how it sees. The very end of almost every image processing system is the human eye. Although our visual system is by far the best image processing system öne can think of, it is also far from being perfect. If the coding scheme is matched to the human visual system and tries to imitate its functions, high compressions can be expected. A few methods designed with these approaches give encouraging >+l) _ -(") " dE T duitj is the relaxation parameter and w G [0, 2]. n denotes the iteration number in increasing order. By taking derivatives and rearranging we obtain, u\n7L> - An) - A2 (141 [4wt-,j - (ui-ij + ui+ltj + Uij-i. + Uitj+i)} where /3ij = 0, and *4j = ^»'.i where fcj = 1. During the SOR iterations, pre-existing data should not be modified and iterative calculations must be performed only for non-existing data. If goal is to construct surface from sparse data, it is not important whether the data is at an edge location or in the middle of a smooth region. The important thing is to locate and use available data for reconstruction. The molecule of a membrane has a structure with five elements. This molecule has its center and the four connected components which are only one pixel apart. This means that any molecule cannot contain pixels in different regions separated by edges, each pixel in membrane functional is enough for reconstruction. The method is applied to various synthetic and real images. The compression ratio for syntetic images is approximately 40:1 and for real images approximately 20:1. Xlll ABSTRACT IMAGE COMPRESSION BY USING WEAK MEMBRANE MODEL The digital representation of an image requires a very large number of bits. Pictorial data is produced by sampling in space and quantizing in brightness analog scenes. The sampling step size is usually chosen small enough to avoid interpolation before display and relies on the integration ability of the human visual system. A digital image is an N by N array of integer numbers ör picture elements (pixels). The goal of image coding is to reduce (to compress), as much as possible, the number of bits necessary to represent and reconstruct a faithful duplicate of the original picture. Any data originated from an image, are not random. Adjacent pixels have similar gray values,this is an important spatial correlation. If this correlation is exploited in an appropriate way, the number of N2B bits required for represen¬ tation of a digital image can be reduced. The set of techniques based on this classical view of image coding problem is called "first generation" [1]. By using information theory various alternatives were proposed, these coding techniques reached a compression ratio at about 10 to 1. This certainly does not mean that the upper bound given by the entropy of the source has also been reached. First, the entropy of an image is unknown and depends heavily on the model used, and second, information theory is not customarily used to take into account what the human eye sees and how it sees. The very end of almost every image processing system is the human eye. Although our visual system is by far the best image processing system öne can think of, it is also far from being perfect. If the coding scheme is matched to the human visual system and tries to imitate its functions, high compressions can be expected. A few methods designed with these approaches give encouraging >+l) _ -(") " dE T duitj is the relaxation parameter and w G [0, 2]. n denotes the iteration number in increasing order. By taking derivatives and rearranging we obtain, u\n7L> - An) - A2 (141 [4wt-,j - (ui-ij + ui+ltj + Uij-i. + Uitj+i)} where /3ij = 0, and *4j = ^»'.i where fcj = 1. During the SOR iterations, pre-existing data should not be modified and iterative calculations must be performed only for non-existing data. If goal is to construct surface from sparse data, it is not important whether the data is at an edge location or in the middle of a smooth region. The important thing is to locate and use available data for reconstruction. The molecule of a membrane has a structure with five elements. This molecule has its center and the four connected components which are only one pixel apart. This means that any molecule cannot contain pixels in different regions separated by edges, each pixel in membrane functional is enough for reconstruction. The method is applied to various synthetic and real images. The compression ratio for syntetic images is approximately 40:1 and for real images approximately 20:1.
Açıklama
Tez (Yüksek Lisans) -- İstanbul Teknik Üniversitesi, Fen Bilimleri Enstitüsü, 1995
Anahtar kelimeler
Bilgisayar ve Kontrol, Görüntü sıkıştırma, Computer Science and Control, Image compression