Evolutionary optimized discrete Tchebichef moments for image compression applications

Abstract

Evolutionary optimized coefficients of discrete orthogonal Tchebichef moment transform (TMT) are utilized in this study to ameliorate the quality of the traditional moments-based image compression methods. Most of the existing methods compute moment-transform coefficients for the input image and then select the coefficients sequentially downward to a certain order based on the desired compression ratio. However, the proposed method divides the input image into nonoverlapping square blocks of specific size in order to circumvent the problem of numerical instability and then computes the TMT coefficients for each block. In this work, a real-coded genetic algorithm is employed to optimize the TMT coefficients of each block, which produces reconstructed images of better quality for the desired compression ratio. Here the optimization is carried out by minimizing the mean square error function. Standard test images of two different sizes (128 × 128 and 256 × 256) have been subjected to the proposed compression method for the block sizes (4 × 4 and 8 × 8) in order to assess its performance. The results reveal that the proposed real-coded genetic algorithm-based method outperforms others, namely the conventional sequential selection method and simple random optimization method, for the chosen input images in terms of the task of compression.

A Machine Learning Algorithm for Biomedical Images Compression Using Orthogonal Transforms

“…Now, let consider a function with two variables   , f x y sampled on a square grid NxN points. Its DChT is given by (10), [12,[29][30][31]:…”

The multi-resolution capability of Tchebichef moments and its applications to the analysis of fluorescence excitation-emission spectra

“…Among the different kinds of image moments, discrete orthogonal moments inherit these advantages and are able to address the problems, such as information redundancy of geometric moments and image coordinate transformation of continuous orthogonal moments [28]. As a family member of the discrete orthogonal moments, the Tchebichef moments have been applied for image compression applications [29], image watermarking [30], image classification [31] and image fusion [32]. Moreover, in our previous studies, the discrete Tchebichef moments have been successfully applied to quantitative/quantitative analysis on the basis of different kinds of three-dimensional (3D) spectra, such as: (1) the four target compounds in a traditional Chinese medicine named Sanqi Panax Notoginseng have been determined on the basis of HPLC-PAD 3D spectra even in the presence of overlapped peaks, unknown interferences and chromatographic drifts [33]; (2) the three active compounds in Qingrejiedu oral liquid have been quantified based on LC-MS 3D spectra, where some unknown components and noise signals existed [34]; (3) the five metabolites in table wines have been quantified from 1 H NMR selfconstructed (SC-3D) spectra that accompanied by peak shifts, overlapped signals and unknown interferences [35]; (4) the three compounds in mixture samples have been determined on the basis of NIR 3D spectra with temperature effect [36]; and (5) the quality assessment of traditional Chinese medicine using HPLC-PAD 3D spectra was carried out, and more accurate and reliable results have been obtained as compared with traditional analysis method [37].…”