A new image compression method based on comparative thresholding

In recent years, image compression methods are the important research topics, these methods are classified into reversible and irreversiblet techniques. The irreversible techniques or called transformation techniques are the most widely used today. In fact, the transformation methods decompose the image into an orthonormal basis, such as a great number of the coefficients be equal to zero, without significantly impairing the visual aspect of the original image. Decades ago, the wavelet theory has given high contributions in the transformation methods by its nice mathematical properties and using the traditional thresholding as Hard and soft thresholding. However, if the image has a large singularity, we will be obliged to increase the number of non-zero wavelet coefficients in the reconstructed image. So, this situation costs us a lot, especially with color images. Therefore, we can pose the following problem to answer: are there thresholding operators can reduce the wavelet coefficients number independently on the singularity contained in the image?. In this paper, we propose a new method for image compression, called Comparative Thresholding (CT), which is based on constructing new thresholding operators that solve the imposed compression problem. These solution operators can enhance the traditional thresholding in terms of the visual quality of the compressed image. Experiments on a grayscale image using Discrete Wavelet Transform (DWT) with our proposed method (CT), show that CT method provides better performance compared to traditional thresholding in terms of the Peak Signal to Noise Ratio (PSNR) and Compression Ratio (CR).