Natural image understanding using algorithm selection and high-level feedback
04 Feb 2013-Proceedings of SPIE (International Society for Optics and Photonics)-Vol. 8662, pp 96-105
TL;DR: An algorithm selection approach that permits to always use the most appropriate algorithm for the given input image by at first selecting an algorithm based on low level features such as color intensity, histograms, spectral coefficients.
Abstract: Natural Image processing and understanding encompasses hundreds or even thousands of different algorithms.
Each algorithm has a certain peak performance for a particular set of input features and configurations of
the objects/regions of the input image (environment). To obtain the best possible result of processing, we
propose an algorithm selection approach that permits to always use the most appropriate algorithm for the
given input image. This is obtained by at first selecting an algorithm based on low level features such as color
intensity, histograms, spectral coefficients. The resulting high level image description is then analyzed for logical
inconsistencies (contradictions) that are then used to refine the selection of the processing elements. The feedback
created from the contradiction information is executed by a Bayesian Network that integrates both the features
and a higher level information selection processes. The selection stops when the high level inconsistencies are all
resolved or no more different algorithms can be selected.
Citations
More filters
01 Jan 2017
TL;DR: The impact of reversible computing and quantum technology strongly modifies the fault types that can appear and thus the fault models that should be considered, which demonstrates what faults can occur and what faults cannot occur in various reversible technologies.
Abstract: In this chapter we describe faults that can occur in reversible circuit as compared to faults that can occur in classical irreversible circuits. Because there are many approaches from classical irreversible circuits that are being adapted to reversible circuits, it is necessary to analyze what faults that exists in irreversible circuits can appear in reversible circuit as well. Thus we focus on comparing faults that can appear in classical circuit technology with faults that can appear in reversible and quantum circuit technology. The comparison is done from the point of view of information reversible and information irreversible circuit technologies. We show that the impact of reversible computing and quantum technology strongly modifies the fault types that can appear and thus the fault models that should be considered. Unlike in the classical non-reversible transistor based circuits, in reversible circuits it is necessary to specify what type of implementation technology is used as different technologies can be affected by different faults. Moreover the level of faults and their analysis must be revised to precisely capture the effects and properties of quantum gates and quantum circuits that share several similarities with reversible circuits. By not doing so the available testing approaches adapted from classical circuits would not be able to properly detect relevant faults. In addition, if the classical faults are directly applied without revision and modifications, the presented testing procedure would be testing for such faults that cannot physically occur in the given implementation of reversible circuits. The observation and analysis of these various faults presented in this chapter clearly demonstrates what faults can occur and what faults cannot occur in various reversible technologies. Consequently the results from this chapter can be used to design more precise tests for reversible logic circuits. Moreover the clearly described differences between faults occurring in reversible and irreversible circuits means that new algorithms for fault detection should be implemented specifically for particular reversible technologies.
8 citations
Proceedings Article•
25 Oct 2013TL;DR: This work proposes a platform based on the algorithm selection approach to the problem of natural image understanding using a bottom-up algorithm selection from real-world image features and a top-down algorithm selection using information obtained from a high level symbolic world description and algorithm suitability.
Abstract: A real-world intelligent system consists of three basic modules: environment recognition, prediction (or estimation), and behavior planning To obtain high quality results in these modules, high speed processing and real time adaptability on a case by case basis are required In each of the above mentioned modules, many different algorithms and algorithms networks exists and provide various performances on a case by case basis Thus, a mechanism that for any of the three computational stages selects the best possible algorithm is required We propose a platform based on the algorithm selection approach to the problem of natural image understanding This selection mechanism is based on machine learning; a bottom-up algorithm selection from real-world image features and a top-down algorithm selection using information obtained from a high level symbolic world description and algorithm suitability To accommodate the high-speed processing requirements, the high-frequency of real-time reconfiguration and a low-cost of implementation, we are using present a novel dynamic reconfigurable VLSI processor for real-time adaptation of the algorithm selection The new architecture includes a finegrain Digital Reconfigurable Processor, a distributed configuration memory to solve the data transfer bottleneck and an intra-chip packet routing scheme to reduce the size of the configuration memory
5 citations
TL;DR: This work proposes an algorithm selection approach that selects the best algorithm for a each input image and shows that the algorithm selected approach is ideally suited for either a hybrid type VLSI processor or for a Logic-In-Memory processing platform.
Abstract: Natural image processing and understanding encompasses hundreds of different algorithms Each algorithm generates best results for a particular set of input features and configurations of the objects/regions in the input image (environment) To obtain the best possible result of processing in a reliable manner, we propose an algorithm selection approach that selects the best algorithm for a each input image The proposed algorithm selection starts by first selecting an algorithm using low level features such as color intensity, histograms, spectral coefficients or so and a user given context if available The resulting high-level image description is analyzed for logical inconsistencies (contradictions) and image regions that must be processed using a different algorithm are selected The high-level description and the optional user-given context are used by a Bayesian Network to estimate the cause of the error in the processing The same Bayesian Network also generates new candidate algorithm for each region containing the contradiction in an iterative manner This iterative selection stops when the high-level inconsistencies are all resolved or no more different algorithms can be selected We also show that when inconsistencies can be detected, our framework is able to improve high-level description when compared with single algorithms In order for such complex and iterative processing being computationally tractable we also introduce a hardware platform based on reconfigurable VLSI that is well suited as the platform of the proposed approach We show that the algorithm selected approach is ideally suited for either a hybrid type VLSI processor or for a Logic-In-Memory processing platform
5 citations
DOI•
31 Mar 2015TL;DR: This paper presents a method of iterative composition of the high level description of the environment recognition module that allows us to select the best result for each region of the image by evaluating all the intermediary representations and finally keep only the best one.
Abstract: A real-world intelligent system consists of three basic modules: environment recognition, prediction (or estimation), and behavior planning. To obtain high quality results in these modules, high speed processing and real time adaptability on a case by case basis are required. In the environment recognition module many different algorithms and algorithm networks exist with varying performance. Thus, a mechanism that selects the best possible algorithm is required. To solve this problem we are using an algorithm selection approach to the problem of natural image understanding. This selection mechanism is based on machine learning; a bottom-up algorithm selection from real-world image features and a top-down algorithm selection using information obtained from a high level symbolic world description and algorithm suitability. The algorithm selection method iterates for each input image until the high-level description cannot be improved anymore. In this paper we present a method of iterative composition of the high level description. This step by step approach allows us to select the best result for each region of the image by evaluating all the intermediary representations and finally keep only the best one.
4 citations
Cites background or methods from "Natural image understanding using a..."
...With respect to general robotic processing [4, 5] the concept of algorithm selection was introduced into the middle and high level processing of natural image segmentation and understanding....
[...]
...This is due to the fact that each of the mechanism in the loop is in generally inaccurate and various combinations of algorithms needs to be tested to obtain a high level representation without contradiction [5]....
[...]
...The input nodes and the structure of the algorithm were obtained after a minimization procedure described in [5]....
[...]
...However in this work our focus is mainly on the diversity of algorithms in the first three level of processing [5]....
[...]
References
More filters
TL;DR: This work treats image segmentation as a graph partitioning problem and proposes a novel global criterion, the normalized cut, for segmenting the graph, which measures both the total dissimilarity between the different groups as well as the total similarity within the groups.
Abstract: We propose a novel approach for solving the perceptual grouping problem in vision. Rather than focusing on local features and their consistencies in the image data, our approach aims at extracting the global impression of an image. We treat image segmentation as a graph partitioning problem and propose a novel global criterion, the normalized cut, for segmenting the graph. The normalized cut criterion measures both the total dissimilarity between the different groups as well as the total similarity within the groups. We show that an efficient computational technique based on a generalized eigenvalue problem can be used to optimize this criterion. We applied this approach to segmenting static images, as well as motion sequences, and found the results to be very encouraging.
13,789 citations
17 Jun 1997
TL;DR: This work treats image segmentation as a graph partitioning problem and proposes a novel global criterion, the normalized cut, for segmenting the graph, which measures both the total dissimilarity between the different groups as well as the total similarity within the groups.
Abstract: We propose a novel approach for solving the perceptual grouping problem in vision. Rather than focusing on local features and their consistencies in the image data, our approach aims at extracting the global impression of an image. We treat image segmentation as a graph partitioning problem and propose a novel global criterion, the normalized cut, for segmenting the graph. The normalized cut criterion measures both the total dissimilarity between the different groups as well as the total similarity within the groups. We show that an efficient computational technique based on a generalized eigenvalue problem can be used to optimize this criterion. We have applied this approach to segmenting static images and found results very encouraging.
11,827 citations
TL;DR: An object detection system based on mixtures of multiscale deformable part models that is able to represent highly variable object classes and achieves state-of-the-art results in the PASCAL object detection challenges is described.
Abstract: We describe an object detection system based on mixtures of multiscale deformable part models. Our system is able to represent highly variable object classes and achieves state-of-the-art results in the PASCAL object detection challenges. While deformable part models have become quite popular, their value had not been demonstrated on difficult benchmarks such as the PASCAL data sets. Our system relies on new methods for discriminative training with partially labeled data. We combine a margin-sensitive approach for data-mining hard negative examples with a formalism we call latent SVM. A latent SVM is a reformulation of MI--SVM in terms of latent variables. A latent SVM is semiconvex, and the training problem becomes convex once latent information is specified for the positive examples. This leads to an iterative training algorithm that alternates between fixing latent values for positive examples and optimizing the latent SVM objective function.
10,501 citations
07 Jul 2001
TL;DR: In this paper, the authors present a database containing ground truth segmentations produced by humans for images of a wide variety of natural scenes, and define an error measure which quantifies the consistency between segmentations of differing granularities.
Abstract: This paper presents a database containing 'ground truth' segmentations produced by humans for images of a wide variety of natural scenes. We define an error measure which quantifies the consistency between segmentations of differing granularities and find that different human segmentations of the same image are highly consistent. Use of this dataset is demonstrated in two applications: (1) evaluating the performance of segmentation algorithms and (2) measuring probability distributions associated with Gestalt grouping factors as well as statistics of image region properties.
6,505 citations
TL;DR: A review of the Pascal Visual Object Classes challenge from 2008-2012 and an appraisal of the aspects of the challenge that worked well, and those that could be improved in future challenges.
Abstract: The Pascal Visual Object Classes (VOC) challenge consists of two components: (i) a publicly available dataset of images together with ground truth annotation and standardised evaluation software; and (ii) an annual competition and workshop. There are five challenges: classification, detection, segmentation, action classification, and person layout. In this paper we provide a review of the challenge from 2008---2012. The paper is intended for two audiences: algorithm designers, researchers who want to see what the state of the art is, as measured by performance on the VOC datasets, along with the limitations and weak points of the current generation of algorithms; and, challenge designers, who want to see what we as organisers have learnt from the process and our recommendations for the organisation of future challenges. To analyse the performance of submitted algorithms on the VOC datasets we introduce a number of novel evaluation methods: a bootstrapping method for determining whether differences in the performance of two algorithms are significant or not; a normalised average precision so that performance can be compared across classes with different proportions of positive instances; a clustering method for visualising the performance across multiple algorithms so that the hard and easy images can be identified; and the use of a joint classifier over the submitted algorithms in order to measure their complementarity and combined performance. We also analyse the community's progress through time using the methods of Hoiem et al. (Proceedings of European Conference on Computer Vision, 2012) to identify the types of occurring errors. We conclude the paper with an appraisal of the aspects of the challenge that worked well, and those that could be improved in future challenges.
6,061 citations