Jonathan J. Ashley

Query by image and video content: the QBIC system

Research on ways to extend and improve query methods for image databases is widespread. We have developed the QBIC (Query by Image Content) system to explore content-based retrieval methods. QBIC allows queries on large image and video databases based on example images, user-constructed sketches and drawings, selected color and texture patterns, camera and object motion, and other graphical information. Two key properties of QBIC are (1) its use of image and video content-computable properties of color, texture, shape and motion of images, videos and their objects-in the queries, and (2) its graphical query language, in which queries are posed by drawing, selecting and other graphical means. This article describes the QBIC system and demonstrates its query capabilities. QBIC technology is part of several IBM products. >

Modulation coding for pixel-matched holographic data storage

We describe a digital holographic storage system for the study of noise sources and the evaluation of modulation and error-correction codes. A precision zoom lens and Fourier transform optics provide pixel-to-pixel matching between any input spatial light modulator and output CCD array over magnifications from 0.8 to 3. Holograms are angle multiplexed in LiNbO(3):Fe by use of the 90 degrees geometry, and reconstructions are detected with a 60-frame/s CCD camera. Modulation codes developed on this platform permit image transmission down to signal levels of ~2000 photons per ON camera pixel, at raw bit-error rates (BERs) of better than 10(-5). Using an 8-12-pixel modulation code, we have stored and retrieved 1200 holograms (each with 45,600 user bits) without error, for a raw BER of <2x10(-8).

The query by image content (QBIC) system

QBIC (Query By Image Content) is a prototype software system for image retrieval developed at the IBM Almaden Research Center. It allows a user to query an image collection using features of image content – colors, textures, shapes, locations, and layout of images and image objects. For example, a user can query for images with a green background that contain a round red object in the upper left. The queries are formed graphically a query for red objects can be specified by selecting the color red from a color wheel, a texture query can be specified by selecting from a palette of textures, a query for a shape can be specified by drawing the shape on a” blackboard”, and so on. Retrievals are based on similarity, not exact match, computed from nuHarry Road, San Jose, CA 95120

Optimal binary index assignments for a class of equiprobable scalar and vector quantizers

The problem of scalar and vector quantization in conjunction with a noisy binary symmetric channel is considered. The issue is the assignment of the shortest possible distinct binary sequences to quantization levels or vectors so as to minimize the mean-squared error caused by channel errors. By formulating the assignment as a matrix (or vector in the scalar case) and showing that the mean-squared error due to channel errors is determined by the projections of its columns onto the eigenspaces of the multidimensional channel transition matrix, a class of source/quantizer pairs is identified for which the optimal index assignment has a simple and natural form. Among other things, this provides a simpler and more accessible proof of the result of Crimmins et al. (1969) that the natural binary code is an optimal index assignment for the uniform scalar quantizer and uniform source. It also provides a potentially useful approach to further developments in source-channel coding.

A note on the Shannon capacity of run-length-limited codes (Corresp.)

It is proven that 100-percent efficient fixed-rate codes for run-length-limited (RLL) (d,k) and RLL charge-constrained (d, k; c) channels are possible in only two eases, namely (d,k; c)=(0,1;1) and (1,3;3) . Specifically, the binary Shannon capacity of RLL (d, k) constrained systems is shown to be irrational for all values of (d, k),0 \leq d . For RLL charge-constrained systems with parameters (d, k;c) , the binary capacity is irrational for all values of (d, k; c),0 \leq d , except (0,1; 1) and (1,3;3) , which both have binary capacity 1/2 .

A 550 Mb/s radix-4 bit-level pipelined 16-state 0.25-/spl mu/m CMOS Viterbi decoder

In todays high-speed disk drive read channel ICs maximum likelihood detection using the Viterbi algorithm is a key component in reconstructing digital data sequences. The presented Viterbi decoder was realized in a 0.25 /spl mu/m CMOS technology. Using the proposed comparison approach, it achieves a throughput rate of 550 Mb/s.

Two-dimensional low-pass filtering codes

We describe a framework for designing encoders that transform arbitrary data sequences into two-dimensional arrays satisfying certain constraints, in particular, constraints that guarantee arrays with limited high spatial frequency content. We also exhibit specific codes that produce such arrays. Such codes are useful for holographic recording systems.

Canonical Encoders for Sliding Block Decoders

The existence and uniqueness of a canonical minimal encoder for any given sliding block decoder are proven. The structure of this encoder is given in terms of an explicit sequence of state splittings. The universality of the state splitting algorithm for code construction is clarified.

A linear bound for sliding-block decoder window size. II

For pt.I see ibid., vol.34, p. 389-99, 1988. An input-constrained channel is the set S of finite sequences of symbols generated by the walks on a labeled finite directed graph G (which is said to present S). We introduce a new construction of finite-state encoders for input-constrained channels. The construction is a hybrid of the state-splitting technique of Adler, Coppersmith, and Hassner (1983) and the stethering technique of Ashley, Marcus, and Roth (see ibid., vol.41, p.55-76, 1995). When S has finite memory, and p and g are integers where p/g is at most the capacity of S, the construction guarantees an encoder at rate p:g and having a sliding-block decoder (literally at rate q:p) with look-ahead that is linear in the number of states of the smallest graph G presenting S. This contrasts with previous constructions. The straight Adler, Coppersmith, and Hassner construction provides an encoder having a sliding-block decoder at rate q:p, but the best proven upper bound on the decoder look-ahead is exponential in the number of states of G. A previous construction of Ashley provides an encoder having a sliding-block decoder whose look-ahead has been proven to be linear in the number of states of G, but the decoding is at rate tq:tp, where t is linear in the number of states of G.

Complexity and sliding-block decodability

A constrained system, or sofic system, S is the set of symbol strings generated by the finite-length paths through a finite labeled, directed graph. Karabed and Marcus (1988), extending the work of Adler, Coppersmith, and Hassner (1983), used the technique of state-splitting to prove the existence of a noncatastrophic, rate p:q finite-state encoder from binary data into S for any input word length p and codeword length q satisfying p/q/spl les/cap(S), the Shannon (1948) capacity. For constrained systems that are almost-finite-type, they further proved the existence of encoders enjoying a stronger form of decodability-namely, sliding-block decodability. In particular, their result implies the existence of a 100% efficient (rate 1/2), sliding-block code for the charge-constrained, runlength-limited constraint with parameters (d, k; c)=(1,3; 3), an almost-finite-type system with capacity precisely 1/2. We describe two quite different constructions of such codes. The constructions highlight connections between the problem of determining sliding-block decodability of a finite-state encoder and certain problems of colorability for graphs and sets. Using these connections, we show that the problem of determining the existence of a block-decodable input tag assignment for a given rate p:q, finite-state encoder is NP-complete, for p>1. We also prove NP-completeness results for several related problems in combinatorics and coding.