Experimental rate distortion analysis of image coder with contourlet transform and SPIHT algorithm

We analyze the results of a series of experiments to investigate the empirical rate distortion performance of a contourlet image coder. We use a contourlet transform with different number of directional decomposition levels, with wavelet transform as a reference; coefficients are coded with the SPIHT algorithm. The experimental setting uses popular test images (‘Lena’, ‘Peppers’ etc) and broad spectrum of quality measures (MSE, PSNR, measures dedicated to quantization degradation and structural similarity). We conclude that within the range of analyzed parameters contourlet transform outperforms wavelets in terms of quality, but because of high computational cost, the gains may not easily translate into overall increased performance.