Image recognition in single-scale and multiscale decoders

At best, evaluation of image coders using PSNR is of questionable perceptual validity. But for several types of algorithms, including those with spatially scalable decoders, PSNR might not even be computable, and other methods of evaluation must be used. With a simple reordering of the transmitted bit stream, the SPIHT algorithm can be made spatially scalable without any loss in performance or in progressivity. We present experimental results comparing this multiscale SPIHT (MSPIHT) against SPIHT in terms of the bit rates at which viewers recognize objects in the reconstructed images. MSPIHT is also compared with SPIHT images which have been downsampled to the same scale as the MSPIHT images. We show that viewers are able to recognize reduced-scale images, such as those compressed by MSPIHT, substantially earlier than images compressed by SPIHT.