Edge-enhanced SPIHT Image Scalable Coding Algorithm

With the continuous development of multimedia technology and its applications,there is a high demand for image coding technology.The scalable image coding technology can decode a part of the code stream according to the actual spatial image resolution,network bandwidth or terminal computing capability,so as to meet the application requirements of image progressive transmission,multi-quality service and image database browsing.This paper focuses on the edge-enhanced image scalable coding algorithm based on multilevel tree set splitting(SPIHT).Firstly,based on the introduction of image coding principle,according to the different quantization scanning strategies,the existing image scalable coding is divided into a tree structure based coding algorithm,a block structure based coding algorithm,a combined tree structure and a block structure.The coding algorithm discusses the main ideas of the three types of coding algorithms.At the same time,the basic principle of SPIHT coding algorithm is introduced,and the research status and shortcomings of SPIHT algorithm are expounded from three aspects:transformation mode,quantization mode and coefficient scanning organization.Secondly,in view of the existing methods at the low code rate,the edges of the image are blurred and the ringing effect is obvious.This paper finds that the wavelet transform has energy leakage between sub-bands,and analyzes the reconstruction precision and high frequency of each sub-band coefficient.The conditional Shannon entropy of the low-frequency sub-bands concludes that the energy leakage is mainly concentrated in the edge,contour or texture area.Furthermore,the high-frequency energy contained in the low-frequency sub-band is extracted by the high-pass filter for repairing the edge coefficient corresponding to the high-frequency coefficient,and the decoding precision of the high-frequency coefficient is improved;and each high-frequency sub-band is performed along its main gradient direction.Sub-transformation increases the amplitude of the important high-frequency coefficients to give priority decoding.Based on this,a SPIHT algorithm based on energy leakage and high frequency enhancement is proposed.The experimental results show that the decoded image of the proposed algorithm has richer edge and texture details when the code rate is 0.01bpp?0.5bpp.The Peak Signal-to-Noise Ratio(PSNR)is compared with the traditional SPIHT algorithm.The average is 3.11dB higher,which is 2.11dB higher than the improved SPIHT algorithm based on the direction space tree.Thirdly,the existing methods are not able to fully explore the correlation between singular points of wavelet transform coefficients,singular points and non-singular points,and introduce signal reconstruction based on wavelet modulus maxima,and give a model based on Maximum value image restoration method.In order to determine the optimal subband of modulus maxima reconstruction,this paper divides the subbands at the end of decoding into three states:accurate decoding,accurate decoding,and inaccurate decoding according to the distribution characteristics of decoding coefficients.For the research method,the relationship between the decoding termination bit plane level,the subband state and the reconstructed PSNR of the SPIHT algorithm at different code rates is analyzed,and it is found that the highest frequency subband is based on the code rate higher than 0.2bpp.The restoration of the modulus maxima usually achieves a higher quality of reconstruction.Based on this,a SPIHT image coding algorithm based on wavelet coefficient modulus maxima reconstruction is proposed.The experimental results show that the decoded image of the proposed algorithm has richer edge and texture details when the code rate is 0.01bpp?0.5bpp.The average SPIHT algorithm for peak signal-to-noise comparison is 2.25dB higher than the medium and low code.The average SPIHT algorithm based on the direction space tree is increased by 2.40 dB on average,and the average is higher than the SPACS algorithm by 0.48 dB at medium and high code rates.At the same time,the algorithm can achieve a significantly higher subjective decoding quality than the existing methods when the code rate is higher than 0.2bpp.