| Near-lossless image compression is a technology that restricts the maximumdistortion of each pixel while maintains a relatively high compression ratio. With thedevelopment of medical imaging, satellite communications, remote sensing, andfingerprint recognition technology, near lossless compression attracts the attention ofmore and more researchers. In this thesis, a near-lossless image compression algorithmis presented based on the existing methods. The algorithm makes use of thecontext-model-based prediction error correction together with the trellis codedquantization (TCQ) and produces better compression results than some referencedalgorithms.The content of the thesis can be summarized as follows:1 From the point of view of conditional entropy, the possibility to obtain the bettercompression efficiencyis analyzed.2 Start with the usual uniform quantization, The trellis coded quantization is discussedbriefly.3 A near-lossless coding system is constructed by combing the prediction errorcorrection and the TCQ. The main ideas of this system are as follows: Firstly, wepredict the value of a pixel in the image by using the method of gradient adaptivemethod. Secondly, we use the context-based error correction approach to correct theprediction error by making use of the local texture of the current pixel and themagnitude of the current prediction error. That means, for a given local context of thecurrent pixel, if the prediction error is always too large or too small, we can calculatethe mean prediction error under this context. If the context comes up again, we cancorrect the corresponding prediction error by subtracting from it the mean predictionerror calculated previously. In this way, the average prediction error can be reduced. Thecorrected prediction error is then quantized by TCQ and the quantization indices areentropy coded to obtain the final compression output.Compared with the methods reported recently, this method yields satisfactory results. |
PDF Full Text Request