[
{
"id": "thesis:6252",
"collection": "thesis",
"collection_id": "6252",
"cite_using_url": "https://resolver.caltech.edu/CaltechTHESIS:02222011-105108216",
"primary_object_url": {
"basename": "Mese_m_2001.pdf",
"content": "final",
"filesize": 44724227,
"license": "other",
"mime_type": "application/pdf",
"url": "/6252/1/Mese_m_2001.pdf",
"version": "v5.0.0"
},
"type": "thesis",
"title": "Image halftoning and inverse reconstruction problems with considerations to image watermarking",
"author": [
{
"family_name": "Me\u015fe",
"given_name": "Murat",
"clpid": "Me\u015fe-M"
}
],
"thesis_advisor": [
{
"family_name": "Vaidyanathan",
"given_name": "P. P.",
"clpid": "Vaidyanathan-P-P"
}
],
"thesis_committee": [
{
"family_name": "McEliece",
"given_name": "Robert J.",
"clpid": "McEliece-R-J"
},
{
"family_name": "Divsalar",
"given_name": "Dariush",
"clpid": "Divsalar-D"
},
{
"family_name": "Franklin",
"given_name": "Joel N.",
"clpid": "Franklin-J-N"
},
{
"family_name": "Perona",
"given_name": "Pietro",
"clpid": "Perona-P"
},
{
"family_name": "Djokovic",
"given_name": "Igor",
"clpid": "Djokovic-I"
}
],
"local_group": [
{
"literal": "div_eng"
}
],
"abstract": "In this work, we first discuss the image halftoning problem. Halftoning is the rendition of\r\ncontinuous-tone pictures on displays that are capable of producing only two levels. There\r\nare several well-known algorithms for half toning. The dot diffusion method for digital\r\nhalf toning has the advantage of pixel-level parallelism unlike the error diffusion method,\r\nwhich is a popular half toning method. The image quality offered by error diffusion is\r\nstill regarded as superior to most of the other known methods. We show how the image\r\nquality obtained using the dot diffusion method can be improved by optimization of the\r\nso-called class matrix. By taking the human visual characteristics into account we show\r\nthat such optimization consistently results in images comparable to error diffusion, without\r\nsacrificing the pixel-level parallelism. The dot diffusion algorithm will be discussed and\r\nby modifying the algorithm, embedded multiresolution property will be added. Later,\r\nwe introduce LUT (Look Up Table) based half toning and tree-structured LUT (TLUT)\r\nhalftoning. We demonstrate how error diffusion characteristics can be achieved with this\r\nmethod. Afterwards, our algorithm will be trained on halftones obtained by Direct Binary\r\nSearch (DBS) which is an algorithm with high computational complexity. The complexity\r\nof TLUT halftoning is higher than that of error diffusion but much lower than that of the\r\nDBS algorithm. Thus, halftone image quality between that of error diffusion and DBS will\r\nbe achieved depending on the size of tree structure in TLUT algorithm.\r\nWe also discuss the inverse halftoning problem. Inverse halftoning is the reconstruction\r\nof a continuous tone image from its halftoned version. We propose two methods for\r\ninverse half toning of dot diffused images. The first one uses Projection Onto Convex Sets\r\n(POCS) and the second one uses wavelets. We then propose a novel and fast method for\r\ninverse halftoning called the Look Up Table (LUT) Method. The LUT for inverse halftoning\r\nis obtained from the histogram gathered from a few sample halftone images and the\r\ncorresponding original images. For each pixel, the algorithm looks at the pixel's neighborhood\r\n(template) and depending upon the distribution of pixels in the template, it assigns\r\na contone value from a precomputed LUT. The method is extremely fast (no filtering is\r\nrequired) and the image quality achieved is comparable to the best methods known for inverse halftoning. The LUT inverse half toning method does not depend on the specific\r\nproperties of the half toning method, and can be applied to any method. An algorithm for\r\ntemplate selection for LUT inverse half toning is introduced. We also extend LUT inverse\r\nhalftoning to color halftones.\r\nThe next topic is image watermarking and effects of halftoning on watermarked images.\r\nWatermarking is the process of embedding a secret signal into a host signal in order to\r\nverify ownership or authenticity. We discuss the effects of applying inverse half toning before\r\ndetection of watermark in half toned images and offer methods to improve watermark\r\ndetection from halftoned images.\r\nFinally, we consider the optimal histogram modification with MSE metric and optimal\r\ncodebook selection problem. Watermarking with histogram modification is one of the few\r\nwatermarking methods which is robust to rotation and scaling. We formulate histogram\r\nmodification problem as finding a transformation such that the error between the input and\r\nthe output signal is minimized and the output signal has the desired histogram. It turns\r\nout that this problem is equivalent to the integer linear programming problem. Then, we\r\nformulate the problem of finding the optimal code book where the codewords can come from\r\na finite set. The equivalent problem again turns out to be a linear integer programming\r\nproblem and the solution is guaranteed to be globally optimal.",
"doi": "10.7907/561b-d051",
"publication_date": "2001",
"thesis_type": "phd",
"thesis_year": "2001"
}
]