Nikola Simic

Coding algorithm for grayscale images based on Linear Prediction and dual mode quantization

We propose a new method of grayscale image compression.The algorithm is based on a prediction technique.We use quantizers designed for discrete input samples.Proposed model provides higher PSQNR up to 6.14dB in comparison to other similar models. This paper proposes a novel algorithm for grayscale image compression based on dual mode quantization that is supported by improved linear prediction scheme. The idea of dual mode quantization comes from desire to exploit advantages of the both uniform and piecewise uniform quantizers, designed for discrete input samples. The algorithm performs quantizers with a low and medium number of quantization levels and with a fixed codeword length by using a pixel value prediction in preprocessing. The correlation of adjacent pixels is exploited as the main idea for improving the quality of image compression. The proposed prediction is linear and very simple for practical realization. An analysis of reconstructed image quality is presented considering several parameters and by comparing with few other methods - BTC, DPCM and with methods that use transformation coding. Experiments are done applying the proposed compression model to several standard grayscale test images. Special attention is given to determination of thresholds values that determine whether and which of the two offered quantizers to use. Moreover, method for determining the value of proposed quantizers variance is explained. Obtained results show that proposed model ensures gain up to 6.14 dB compared to the BTC model that uses fixed piecewise uniform quantization for discrete input without a pixel value prediction as well as gain up to 5.89 dB compared to the DPCM model that applies dual predictor. The proposed algorithm could find application in current grayscale image compression and video standards.

Image coding algorithm based on Hadamard transform and simple vector quantization

Transform coding is commonly used in image processing algorithms to provide high compression ratios, often at the expense of processing time and simplicity of the system. We have recently proposed a pixel value prediction scheme in order to exploit adjacent pixel correlation, providing a low-complexity model for image coding. However, the proposed model was unable to reach high compression ratios retaining high quality of reconstructed image at the same time. In this paper we propose a new segmentation algorithm which further utilizes adjacent pixel correlation, provides higher compression ratios and it is based on application of Hadamard transform coding. Additional compression is provided by using vector quantization for a low number of quantization levels and by simplifying generalized Lloyd’s algorithm where the special attention is paid to determination of optimal partitions for vector quantization, making a fixed quantizer. The proposed method is quite simple and experimental results show that it ensures better or similar rate-distortion ratio for very low bit-rates, comparing to the other similar methods that are based on wavelet or curvelet transform coding and support or core vector machine application. Furthermore, the proposed method requires very low processing time since the proposed quantizers are fixed, much less than the required time for the aforementioned methods that we compare with as well as much less than the time required for fractal image coding. In the end, the appropriate discussion is provided comparing the results with a scheme based on linear prediction and dual-mode quantization.

Performance of Quasi-logarithmic Quantizer for Discrete Input Signal

In this paper, performance of quasi-logarithmic quantizer, designed for correlated discrete input signal is analyzed. Quantizer design is done for Laplacian source due to its both hardware and software significance, whereas experiments are done by processing test wideband speech signal sampled at 16 [kHz]. The quantizer is exploited as a second stage of two-stage quantization system, where the first step is used for continuous signal sampling, while the second stage provides additional data compression. The main goal is to provide improved design by discussing theoretical performance of two quantization models. As the traditional models for performance estimation provide estimation of average performance, we have decided to propose a novel model for performance estimation and to analyze performance in details for each random input signal variance. Finally, the experimental results have shown excellent matching with theoretical results. DOI: http://dx.doi.org/10.5755/j01.itc.46.3.16197

Performance analysis of companding quantizer using piecewise linear approximation of the Gaussian source

In this paper a piecewise-linear approximation of a probability density function is performed. For approximated probability density function an optimal compressor function is determined. Based on compressor function determined in this way, companding scalar quantizer is designed. Performance analysis is performed for the Gaussian source at the entrance of the quantizer. Obtained results show that the proposed model has performance very close to that of the optimal nonlinear compandor. Furthermore, practical implementation of proposed model does not require integral equations solving. Consequently, the proposed model is much more suitable for implementation.

Analysis of the impact of the quantizer range choice on compression and quality of the reconstructed image

In this paper an algorithm for grayscale image compression based on usage of three fixed uniform quantizers designed for discrete input samples is presented. The algorithm is based on the alternating use of these three quantizers. Number of quantization levels and quantizer range size increases from the first to the third quantizer. Experimental results show that choice of the quantizer range has an impact on system performance. While selecting a range of the first two quantizers (with a lower number of quantization levels) it is necessary to make a compromise between quality and bit rate (larger quantizer range leads to lower average bit rate but the quality of reconstructed image is also lower). It is shown that the range of the third quantizer should be set up to cover as many as possible high number of input samples making sure that the overload distortion does not become dominant. [Projekat Ministarstva nauke Republike Srbije, br. III 044006]

Design of quantizer for Laplacian source based on the probability density function approximation

In this paper, piecewise linear approximation of the probability density function is done. On the basis of the obtained approximated probability density function, a companding quantizer is designed. The signal at the entrance of the companding quantizer is modeled by Laplacian probability density function. Performances of the proposed companding quantizer are estimated on the basis of comparison of calculated values of signal to quantization noise ratio and approximation error, with the calculated values that correspond to the model of piecewise uniform scalar quantizer, that is also proposed in this paper.