Summary: | 碩士 === 國立中興大學 === 電機工程學系所 === 104 === Direct binary search can produce the best quality halftone images among the current available halftone algorithms. Its method is based on a perceptual filter which is generated from the perceptual function of the human visual system. With the perceptual filter, it searches for the halftone output which produces the minimal squared perceptual error. Therefore, the selection of the perceptual filter will affect the final halftone image. For this reason, the clustered-dot direct binary search (CLU-DBS) uses two different filters in the initial and update stages to generate a difference filter with bandpass characteristic and produces a clustered-dot halftone image.
In the original CLU-DBS researches, the effects of the initial and update filters in the halftone images are not thoroughly investigated. Therefore, in this thesis, we study the effects of the initial filter and two kinds of update filters in CLU-DBS. The first kind of update filter is Gaussian update filter. In association with Gaussian initial filter, the difference filter is a bandpass filter with a single passband. The second kind of update filter is a new type of filter called “multiple-band update filter” (MBU filter). Using this new update filter in CLU-DBS, the difference filter between the update and initial filters exhibits two passbands. Therefore, we can generate halftone images with two types of clusters with different characteristics and create clustered-dot images that are different from those using Gaussian filters.
|