The multitude of visual media-based applications demands sophisticated compression and indexing techniques for efficient storage, transmission and retrieval of images and video. Compressed domain techniques are becoming popular due to their inherent advantages of efficiency and reduced complexity. Wavelet transform is emerging as a powerful tool for efficient compression of visual information. A variety of wavelet based video compression techniques have been reported in the literature. However, there has been little work done in the area of video indexing in the wavelet domain. This is crucial for consideration of wavelets as a potential candidate for multimedia applications, including standards such as MPEG-7. In this paper, we first present a brief review of waveletbased indexing techniques. We then propose an integrated video coding and indexing system in a wavelet compression framework. In the proposed system, the compression parameters associated with the video coder are employed in all steps of video indexing, namely video segmentation, extraction of temporal indexing (object motion and camera operations) and spatial indexing parameters. Simulation results show that the techniques proposed towards the design of the wavelet based integrated video coding and indexing system provide a good indexing performance at a low complexity.

Traditionally, processing an image stored in the compressed form requires decompression first. Following manipulations, the processed image is recompressed for storage. Manipulating images in the semicompressed or transform domain appears to be an effective way to reduce processing time. In this paper, new algorithms for image interpolation in the Haar transform domain are proposed. These algorithms perform interpolation by computing a new subband using two different approaches. Experimental results indicate that in addition to running faster in the transform domain, the new algorithms produce images with qualities comparable to those in the spatial domain. Applications are seen in conjunction with Haar-based transforms such as the HCOMPRESS which is commonly used for compression in space science imagery.

Motion estimation is a temporal image compression technique, where an nxn block of pixels in the current frame of a video sequence is represented by a motion vector with respect to the best matched block in a search area of the previous frame. In this paper, a low complexity technique for motion estimation is proposed. The proposed technique, first, reduces the n-bit grayscale frames into 1-bit binary frames using morphological filters, and determines the displacement of the edge features of the adjacent frames. While reduction in bit depth requires a small percentage of computation using the full pixel resolution, the search procedure is performed by simple XOR logic operations and 1-b distortion accumulations on the entire search area. Compared to other low complexity techniques, the proposed technique yields better frame reconstruction, operates using simpler arithmetic/logic operations, and possesses a higher degree of parallelism when implemented on a twodimensional single-instruction stream, multiple-data stream architecture.

This paper proposes a novel class of multichannel filters called neuro-fuzzy hybrid multichannel (NFHM) filters to simultaneously achieve three objectives: noise attenuation, chromaticity retention, and edges or details preservation. NFHM filters are characterized by a set of fuzzy rules (structure knowledge) such that they are capable of effectively fusing together the useful filtering merits from vector median, vector directional, and identity filters to further improve the filtering performance of the conventional filters. Moreover, we adequately exploit the functional equivalence between fuzzy inference systems and radial-basis function networks on the optimal design of NFHM filters such that NFHM filters can be optimized by neuro-learning techniques based on the radial-basis function networks to obtain adaptive fuzzy rules for the different window contents. Finally, extensive simulation results demonstrate that the filtering performance of NFHM filters is superior to that of other proposed filters.

The usefulness of the hit–miss transform (HMT) and related transforms for pattern matching in document image applications is examined. Although the HMT is sensitive to the types of noise found in scanned images, including both boundary and random noise, a simple extension, the blur HMT, is relatively robust. The noise immunity of the blur HMT derives from its ability to treat both types of noise together, and to remove them by appropriate dilations. In analogy with the Hausdorff metric for the distance between two sets, metric generalizations for special cases of the blur HMT are derived. Whereas Hausdorff uses both directions of the directed distances between two sets, a metric derived from a special case of the blur HMT uses just one direction of the directed distances between foreground and background components of two sets. For both foreground and background, the template is always the first of the directed sets. A less-restrictive metric generalization, where the disjoint foreground and background components of the template need not be set complements, is also derived. For images with a random component of noise, the blur HMT is sensitive only to the size of the noise, whereas Hausdorff matching is sensitive to its location. It is also shown how these metric functions can be derived from the distance functions of the foreground (FG) and background (BG) of an image, using dilation by the appropriate templates. The blur HMT can be used as a fast heuristic to avoid more expensive integer-based matching techniques, and it is implemented efficiently with boolean image operations. The FG and BG images are dilated with structuring elements that depend on image noise and pattern variability, and the results are then eroded with templates derived from patterns to be matched. Subsampling the patterns on a regular grid can improve speed and maintain match quality, and examples are given that indicate how to explore the parameter space. Truncated matches give the same result as full erosions, are much faster, and for some applications can be performed at a restricted set of locations.

Many binary halftoning algorithms tend to render extreme tones (i.e., very light or very dark tones) with objectionable dot distributions. To alleviate this artifact, we introduce a halftone postprocessing algorithm called the Springs algorithm. The objective of Springs is to rearrange minority pixels in affected regions for a smoother, more attractive rendition. In this paper, we describe the Springs algorithm, and we show results which demonstrate its effectiveness. The heart of this algorithm is a simple dot-rearrangement heuristic which results in a more isotropic dot distribution. The approach is to treat any well-isolated dot as if it were connected to neighboring dots by springs, and to move it to a location where the energy in the springs is a minimum. Applied to the whole image, this could degrade halftone appearance. However, Springs only moves dots in selected regions of the image. Pixels that are not minority pixels are not moved at all. Moreover, dot rearrangement is disabled on and around detected edges, since it could otherwise render those edges soft and diffuse.

A new method has been developed for measuring the similarity of two digital images using a common multiscale framework on luminance and texture features. This method applies a multivalued inhomogeneous diffusion model for luminance and texture features to detect multiscale object boundaries. The orientations of the detected boundary points are utilized to obtain a similarity measure, which is defined by matching the orientation histogram pairs determined for each scale level. By applying normalization and histogram shifting, this measure can also address scale and rotation invariance. The method is evaluated on the original and transformed images of Corel Gallery and Kodak photo-CD data by applying image scaling, rotation, and blurring. A similarity ratio of more than 95% is achieved for the first two transformations, and more than 80% for the third.

We describe a modification of the mixture proportion estimation algorithm based on the granulometric mixing theorem. The modified algorithm is applied to the problem of counting different types of white blood cells in bone marrow images. In principle, the algorithm can be used to count the proportion of cells in each class without explicitly segmenting and classifying them. The direct application of the original algorithm does not converge well for more than two classes. The modified algorithm uses prior statistics to initially segment the mixed pattern spectrum and then applies the oneprimitive estimation algorithm to each initial component. Applying the algorithm to one class at a time results in better convergence. The counts produced by the modified algorithm on six classes of cells—myeloblast, promyelocyte, myelocyte, metamyelocyte, band, and PolyMorphoNuclear (PMN)—are very close to the human expert’s numbers; the deviation of the algorithm counts is similar to the deviation of counts produced by human experts. The important technical contributions are that the modified algorithm uses prior statistics for each shape class in place of prior knowledge of the total number of objects in an image, and it allows for more than one primitive from each class.

A model-based FLIR ATR algorithm is described. It utilizes boundary contrast for target detection and recognition. Boundary contrast is related to the location uncertainty at target boundary points. A polygon model is used for deriving target centroid location uncertainty caused by the boundary point location uncertainty. The significance of the work lies in the sound mathematical models used in deriving the relationship between contrast and location uncertainty for the boundary points and the relationship between boundary point location uncertainty and centroid location uncertainty. Experiment results show significantly improved performance in detection, recognition, and localization.

The optical flow observed by a moving camera satisfies, in the absence of noise, a special equation analogous to the epipolar constraint arising in stereo vision. Computing the ‘‘flow fundamental matrix’’ of this equation is an essential prerequisite to undertaking three-dimensional analysis of the flow. This article presents an optimal formulation of the problem of estimating this matrix under an assumed noise model. This model admits independent Gaussian noise that is not necessarily isotropic or homogeneous. A theoretical bound is derived for the accuracy of the estimate. An algorithm is then devised that employs a technique called renormalization to deliver an estimate and then corrects the estimate so as to satisfy a particular decomposability condition. The algorithm also provides an evaluation of the reliability of the estimate. Epipoles and their associated reliabilities are computed in both simulated and real-image experiments. Experiments indicate that the algorithm delivers results in the vicinity of the theoretical accuracy bound.

A technique is described for the detection of defects in digital video images, caused by scratching to the original 2 in. quadruplex video tape. This technique relies upon the reorientation of the digital video data into a manner similar to that in which the original analogue data would have been found on the video tape. This brings the scratched data into alignment and makes the task of detecting the scratched positions a great deal easier. In addition to enabling the detection of scratches in the video data, the technique enables new insight into the mechanisms through which the scratching was first brought about.

Identification of individuals on the basis of facial features is the most natural method of distinguishing one individual from another. In this paper, we propose novel histogram and wavelethistogram methods as means of identifying individuals in a mugshot application. We note that these methods are suitable for real time implementation and provide a high degree of tolerance towards facial expression changes.

Together with the growing interest in the development of human and computer interface and biometric identification, human face recognition has become an active research area since early 1990. Nowadays, principal component analysis (PCA) has been widely adopted as the most promising face recognition algorithm. Yet still, traditional PCA approach has its limitations: poor discriminatory power and large computational load. In view of these limitations, this article proposed a subband approach in using PCA— apply PCA on wavelet subband. Traditionally, to represent the human face, PCA is performed on the whole facial image. In the proposed method, wavelet transform is used to decompose an image into different frequency subbands, and a midrange frequency subband is used for PCA representation. In comparison with the traditional use of PCA, the proposed method gives better recognition accuracy and discriminatory power; further, the proposed method reduces the computational load significantly when the image database is large, with more than 256 training images. This article details the design and implementation of the proposed method, and presents the encouraging experimental results.

This is an edited volume consisting of four parts which include some carefully authored chapters and some selected papers from the literature. It is in between a textbook and a reference book in that it contains several introductory chapters authored by the editor and some advanced research papers on the subject.