Evaluating image segmentation algorithms using monotonic hulls in fitness/cost space

2009 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT)

2009

Automatic parameter selection for image segmentation is accelerated by means of a genetic algorithm (GA). Issues remain in selecting the population size, number of generations, and termination of the search. As evaluation time and subsequent image batch-processing time, when the algorithm is applied in practice, are important considerations, this paper introduces a time factor into the GA cost function. It is found that this procedure while preserving the GA solution also improves interpretation and parameter selection.

“…al. [5]. The analysis of the importance of considering time as a factor in segmentation is intersting.…”

Section: Optimizing Parameters With a Gamentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Time-weighted quantitative testing of image segmentation with a genetic algorithm

2009 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT)

2009

“…Contrasting image segmentation to recognition tasks such as the use of handwriting, and face databases, the authors of [11] remark "Typically [in segmentation] researchers will show their results on a few images and point out why the results 'look good'". Part of the problem is the difficulty of establishing metrics, whether pixel-based figures of merit [12] or rankings against a set of criteria [13] or through precision-recall curves [5]. However, part of the problem may also be the logistics of performing a large number of tests.…”

Section: Related Workmentioning

confidence: 99%

Computationally Efficient Quantitative Testing of Image Segmentation with a Genetic Algorithm

2007 Third International IEEE Conference on Signal-Image Technologies and Internet-Based System

2007

Abstract-Quantitative testing of segmentation algorithms implies rigorous testing against ground truth segmentations. Though under-reported in the literature, the performance of a segmentation algorithm depends on the choice of input parameters. The paper reports wide variety both in evaluation time and segmentation results for an example mean-shift algorithm. When testing extends over an algorithm's parameter space, then the search for satisfactory settings has a considerable cost in time. This paper considers the use of a genetic algorithm (GA) to avoid an exhaustive search. As application of the GA drastically reduces search times, the paper investigates how best to apply the GA in terms of initial candidate population, convergence speed, and application of a final polishing round. The GA parameter search forms part of a three-component computation environment aimed at automating the search and reducing the evaluation time. The first component relies on scripted testing and collation of results. The second component transfers to a commodity cluster computer. And the third component applies a genetic algorithm to avoid an exhaustive search.

“…Contrasting image segmentation to recognition tasks such as the use of handwriting, and face databases, the authors of [3] remark "Typically [in segmentation] researchers will show their results on a few images and point out why the results 'look good'". Part of the problem is the difficulty of establishing metrics, whether pixel-based figures of merit [4] or rankings against a set of criteria [5] or through precision-recall curves [6]. However, part of the problem may also be the logistics of performing a large number of tests.…”

Section: Introductionmentioning

confidence: 99%

A computationally efficient evaluation environment for image segmentation

2007 International Conference on Machine Vision

2007

Abstract-An emphasis on quantitative testing of segmentation algorithms implies rigorous testing against ground truth segmentations. When testing extends over the algorithm's parameter space, then the search for a best fit has a considerable cost in time. The paper reports wide variety both in evaluation time and segmentation results for an example mean-shift algorithm. This paper proposes a three-component computation environment aimed at automating the search and reducing the evaluation time. The first component relies on scripted testing and collation of results. The second component transfers to a commodity cluster computer. And the third component introduces a genetic algorithm to avoid an exhaustive search. Application of the genetic algorithm drastically reduces search times.