Halftone/Contone Conversion Using Neural Network and Its Application to Image Compression

• Main Authors: Yen-Wei Lu, 盧彥瑋
• Other Authors: Wen-Yu Su
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/32410126087045642683

碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 91 === Halftoning [17] technologies are important to an image printing device, be it for color or grey scale images. Dithering and error diffusion [17] are two most popular techniques. However, the performances provided by these methods are mostly evaluated subjectively. On the other hand, inverse halftoning techniques are less researched because the applications are much rarer. However, it is sometimes necessary to convert a halftone image to a contone (Continuous Tone) [17] image when one wants to see it in devices such as computer monitors. In this thesis, a Neural Network based inverse halftoning method is first proposed. This method combines two stages of processing. The first stage is a RBF (Radial Basis Function) network [11] which performs the basic inversion. The second stage is a MLP (Multi-Layer Perceptron) network [11] for post-processing. After inverse halftoning processing, we are able to compare the reconstructed image and the original image using distortion measure such as MSE (Mean Square Error or PSNR). Next, we formulate the halftoning processing as another feedforward neural network based on error diffusion methods. The errors between the original images and the reconstructed images are fed back to train the overall network which includes the two-stage inverse halftoning network and the halftoning network such that both processings achieve the best quality when combined together. Computer simulations show that the proposed schemes provide better visual quality of halftone images and higher PSNR performance compared to other inverse halftoning methods.