Parallel image processing (Q4494541)

The aim of the book under review may to a large extent be derived from the Preface and from Chapter 1, and may be described as follows: a lot of topics in image processing are very well suited for the use of synchronous massively parallel computer systems, which work according to the SIMD principle (Single Instruction, Multiple Data); the authors want to present different topics of lower level image processing which can be implemented efficiently in synchronous parallel algorithms. To this end, the data-parallel programming language Parallaxis (developed by one of the authors) is used, and throughout the book the algorithms are explained and illustrated by giving short parts of a parallel program. The book itself consists of 13 chapters of unequal length (summing up to 168 pages), followed by an extended appendix on the parallel programming language Parallaxis (31 pages) and an index, for a total of 203 pages. The following chapter headings describe the treated topics: Introduction (pp. 1-6); Point operator's (7-17); Local operators (19-26); Edge detection (27-32); Skeletonizing (33-49); Morphological operator (51-58); Segmentation (59-63); Corner detection (65-81); Hough transform (83-97); Fourier transform (99-119); Texture recognition (121-130); Stereo image processing (131-143); Analysis of image sequences (145-168). The stated topics, together with the chapter length, give a good impression of what may be expected from the book, at least to the reviewer's opinion: it is not intended to be a real introductory textbook into the general domain of image processing, but it should rather be seen as an accompanying volume containing a condensed description of some methods but stressing the (parallel) algorithmic-programming facts of those methods. The role of this book as accompanying (or, put another way, complementing) an introductory or advanced text book on the subject may also be concluded from other facts, such as: (i) the presentation does not contain deep mathematical facts, (ii) the book does not contain lists of selected exercises, (iii) the text sometimes supposes (either virtually or explicitly) from the reader some background in the matter, e.g., by referring to existing textbooks or to sequential analogous for algorithms. The number of misprints seems to be very low. To mention just one misprint (for the author's sake), referring to 9.2 and 9.3 on page 67 should really be 8.2 and 8.3; on the same page, the meaning of a sentence would probably become clearer when the word ``be'' should be changed into ``by''. A somewhat more profound problem seems to be incorporated in the use of the Sobel operator: in chapter 4, the Sobel operator (for edge detection) is defined as usual in a \((3\times 3)\)-neighborhood; however, in the same chapter, and once again in chapter 8, use of the Sobel operator in a \((5\times 5)\)-neighborhood is mentioned, without giving its definition. To conclude: although the book under review emphasizes the use of intrinsic parallel algorithms in image processing, it may be a welcome complement to a traditional introductory text book in the domain.