Performance Improvement of Lossy Image Compression Based on Polynomial Curve Fitting and Vector Quantization

Abstract

Lossy image compression performs a fundamental role in modern communication technology to cope up with the transmission and storage problems. In this paper, we present a new efficient lossy image compression method based on the polynomial curve fitting approximation technique, which represents many pixels of the image by a minimum number of polynomial coefficients. The presented method starts by converting the image into one-dimensional signal and it divides this one-dimensional signal into segments of variable length. Then, the polynomial curve fitting is applied to these segments to construct the coefficients matrix. The number of pixels is selected depending on the Root Mean Squared Error threshold value. Finally, the coefficients matrix is quantized using the vector quantization which composed of three procedures: codebook design procedure, encoding procedure, and decoding procedure. The proposed method is implemented for gray and colored images. Experimentally, the proposed method has improved the reconstruction quality by 2–9 dB with a better compression ratio relative to other techniques. Also, the proposed method obtains a better result than any other compared algorithms.