With the growing importance of low-bandwidth applications, such as wireless access to the Internet, images are often sent or received at low bit rates. At these bit rates, they suffer from significant distortion and artifacts, making it difficult for those viewing the images to understand them. We present two progressive compression algorithms that focus on preserving the clarity of important image features, such as edges, at compression ratios of 80:1 and more. Both algorithms capture and encode the locations of important edges in the images. The first algorithm then transmits a standard SPIHT (set partitioning in hierarchical trees) bit stream, and at the decoder applies a nonlinear edge-enhancement procedure to improve the clarity of the encoded edges. The second approach uses a modified wavelet transform to "remove" the edges, and encodes the remaining texture information using SPIHT. With both approaches, features in the images that may be important for recognition are well preserved, even at low bit rates.

Mean squared error (MSE) and peak signal-to-noise-ratio (PSNR) are the most common methods for measuring the quality of compressed images, despite the fact that their inadequacies have long been recognized. Quality for compressed still images is sometimes evaluated using human observers who provide subjective ratings of the images. Both SNR and subjective quality judgments, however, may be inappropriate for evaluating progressive compression methods which are to be used for fast browsing applications. In this paper, we present a novel experimental and statistical framework for comparing progressive coders. The comparisons use response time studies in which human observers view a series of progressive transmissions, and respond to questions about the images as they become recognizable. We describe the framework and use it to compare several well-known algorithms (JPEG, set partitioning in hierarchical trees (SPIHT), and embedded zerotree wavelet (EZW)), and to show that a multiresolution decoding is recognized faster than a single large-scale decoding. Our experiments also show that, for the particular algorithms used, at the same PSNR, global blurriness slows down recognition more than do localized "splotch" artifacts.

Many progressive wavelet-based image coders are designed for good performance on natural images. They attempt to achieve the greatest reduction in mean squared error (MSE) with each bit sent, an approach that is most effective when the image is composed chiefly of low-frequency content. Many images, however, include sharp-edged objects, text characters or graphics that are not well handled by standard wavelet-based methods. These features, which may contain information important for recognition, become distorted and obscured when highly compressed by standard wavelet-based methods. In this paper, we present a new progressive image coder that treats an image as being composed of three types of information: edges, texture, and edge-associated detail. The locations of important edges are encoded using line graphic techniques. Texture is encoded using a wavelet-based zerotree approach. Detail near edges-that cannot be efficiently encoded as texture-is encoded separately with a bitplane coding technique. With this approach, features in the image that may be important for recognition are well preserved, even at low bit rates.

The JBIG2 standard for lossy and lossless bi-level image coding is a very flexible encoding strategy based on pattern matching techniques. The encoder collects a set of symbols in a dictionary and encodes a page by reference to the dictionary symbols. JBIG2 allows the encoder to view all symbols and choose a good set for the dictionary. We propose a two-pass technique for choosing the dictionary entries, and discuss the trade-offs in bit allocation that these choices entail. The proposed dictionary design technique outperforms simpler dictionary formation techniques by up to 7-9% and 10-18% for lossless and lossy compression, respectively. In the lossy case, we further point out one way to significantly speed up the symbol bitmap compression procedure.

Many current progressive wavelet-based image coders attempt to achieve the greatest reduction in mean squared error (MSE) with each bit sent. In so doing, they tend to send information on the lowest-frequency wavelet coefficients first. At very low bit rates, images compressed by these coders are therefore dominated by low frequency information and blotchy artifacts. These effects combine to hamper recognition of objects in the images. In this paper, we present a new progressive image coder which employs edge enhancement with the goal of improving the visual appearance and recognizability of compressed images at very low bit rates. Important edges in the original image are captured and transmitted as side information together with a traditional wavelet coder bit stream. The decoder combines the two complementary information sources in a manner which, for certain image classes, can yield highly recognizable images at very low bit rates.