Back to results list
Please use this identifier to cite or link to this item:
|Title:||On the restoration and formation of color-quantized images||Authors:||Fung, Yik-hing||Keywords:||Hong Kong Polytechnic University -- Dissertations
Image processing -- Digital techniques
|Issue Date:||2006||Publisher:||The Hong Kong Polytechnic University||Abstract:||Color quantization is an image processing technique which reduces the number of colors used to represent an image at a minimum quality penalty. It has been widely used in a number of applications such as displaying an image in a low-end display, delivering images over the Internet and producing hardcopies of color images. To a certain extent, color quantization can be considered as a lossy compression process in which bits per pixel are reduced. A color-quantized image cannot be perfectly reconstructed so as to produce its original version. The restoration of a color-quantized image is hence necessary. Image restoration is a field in which people concern the reconstruction or estimation of an uncorrupted image from its distorted version. Conventional restoration algorithms are dedicated for restoring noisy blurred images. Direct application of existing restoration techniques are generally inadequate to deal with the restoration of compressed images since the degradation models of the two cases are completely different. Though some dedicated algorithms for restoring JPEG-encoded images and halftoned images have been proposed recently, little effort has been seen in the literature for restoring color-quantized images. This thesis presents four novel algorithms for restoring color-quantized images. When an image is color-quantized, a palette is used to define the available colors that can be produced in the output. The smaller the palette size, the more artefacts are introduced in the output. Halftoning is an image processing technique which reduces these artefacts by making use of the property of our human visual system. In the four presented restoration algorithms, two of them were proposed for restoring color-quantized images in which no error diffusion are involved and the other two handle the case when error diffusion is involved. These algorithms were developed independently by tackling the technical problems with different techniques including Regularization, Projection onto Convex Sets (POCS) and Simulated Annealing (SA) separately. All these algorithms make a good use of the available color palette to derive useful a priori information for restoration. As mentioned before, color quantization is commonly used in printing applications to produce high quality hardcopies of color images. In this particular application, without lose of generality, a full color image is decomposed into three color planes and each plane is halftoned with binary halftoning algorithm independently. Multiscale error diffusion is a recently proposed halftoning technique. This technique was known to be superior to conventional halftoning techniques such as error diffusion by eliminating directional hysteresis completely. To evaluate its performance and explore its application in producing color hardcopies, a detailed analysis on multiscale error diffusion was carried out. This thesis presents a report of our analysis.
In order to efficiently render halftone images for various printers and displays which support different resolutions, it is desirable that halftoning results of different scales can be produced at a time and all of them can be embedded in a single full-scale halftone image such that a simple down-sampling process can extract images of suitable resolutions from the full-scale halftone image if necessary. This scalable property was addressed in this thesis. Based on the aforementioned analysis on multiscale error diffusion, we extended its idea and proposed a binary scalable multiresolution halftoning algorithm. This algorithm can produce scalable color prints by handling different color planes separately. With the rapidly evolving computer and communication technologies, the Internet has become the most popular media for information exchange among remote sites throughout the world. There are many forms of information available over the Internet today and digital color images are very popular among them. Images are compressed before its delivery over the Internet, which results in different output image file formats. GIF format is a popular image format generated with color quantization. Since networks of different channel bandwidth and clients of different capability scattered over the Internet, it is desirable to make the color-quantized materials scalable so as to save the channel bandwidth and other resources. A multiscale vector error diffusion algorithm for color quantization was proposed in this thesis. Unlike those halftoning algorithms for printing color images, the proposed algorithm does not handle color planes separately and is able to handle color quantization using any arbitrary palettes. This proposed algorithm was used as a framework for generating scalable color-indexed images and a multiscale multiresolution vector error diffusion algorithm was proposed accordingly. Images possessing this scalable property support transmission over the Internet which contains clients with different display resolutions, system with different caching resources and networks with varying bandwidths and QoS capabilities.
|Description:||xxix, 173 leaves : ill. (some col.) ; 30 cm.
PolyU Library Call No.: [THS] LG51 .H577P EIE 2006 Fung
|URI:||http://hdl.handle.net/10397/3508||Rights:||All rights reserved.|
|Appears in Collections:||Thesis|
Show full item record
Files in This Item:
|b19736344_link.htm||For PolyU Users||160 B||HTML||View/Open|
|b19736344_ir.pdf||For All Users (Non-printable)||4.5 MB||Adobe PDF||View/Open|
Citations as of Mar 12, 2018
Citations as of Mar 12, 2018
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.