C. Brice and C. Fennema, "Scene analysis using regions," Art. Intell., vol. 1, pp. 205--226, 1970.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....properties of the resulting representations. However, it is somewhat laborious as a practical algorithm. In practice, the Crack Insertion Algorithm is the easiest way to derive a topological representation. In similar form, phase 1 of the algorithm has been used in the segmentation methods of [2, 3, 17]. However, phase 2 (derivation of an explicit topological representation) was only carried out in [2] The algorithm starts from a region image, i.e. a complete image partitioning into 4 connected components. Region images typically result from region growing (e.g. 1] split and merge (e.g. ....

....into the data representation (phase 1) After this an XPMap representation can easily be derived (phase 2) Compare figure 4 for illustration: Phase 1: Crack Insertion Given: Region image with labeled 4 connected components, size w x h. 1. Create an image of size (2w 1) x (2h 1) Following [3], we will call this image the super grid. Copy the labels from position (x, y) of the region image to position (2x, 2y) in the super grid. 2. For each call in the super grid with coordinates (2m, 2n 1) If the cell s two vertical neighbors have the same label, copy this label into the cell. ....

C. Brice, C. Fennema: "Scene Analysis Using Regions", Artificial Intelligence, 1(3), pp. 205-226, 1970

....regions of the image satisfy a given homogeneity criterion. Among region segmentation methods, three main different approaches may be distinguished: i) The top down approach [21,19,8] starts with large regions and divide the image into smaller and smaller regions. ii) The bottom up approach [5,14] starts with small regions and merges them until the homogeneity criterion is satisfied. iii) The mixed approach [6,22,9] introduced by Pavlidis and Horowitz [15,16] modifies the partition by splitting or merging regions until they satisfy some homogeneity criterion. Horowitz and Pavlidis have ....

R. Brice and C.L. Fennema. Scene analysis using regions. Artificial intelligence, 1:205--226, 1970.

....than otherwise possible. 2.4 Scene analysis studies Outdoor scene analysis is a complex problem. A number of different approaches have been used for recognising different objects in such scenes. In early experiments on scene analysis, simple problems were tackled. For example Brice and Fennema[24] defined the procedure to interpret simple objects in images such as wedges, cubes, wall and floor. They define a simple procedure for grouping regions and understanding them to see their shape properties for object recognition. Also semantic information, such as the fact that there is a specific ....

C.R. Brice and C.L. Fennema, Scene analysis using regions, Artificial Intelligence, vol. 1, pp. 205-226, 1970.

.... the exhaustive survey of variational models in image processing is beyond the scope of this paper, we defer to the much more complete bibliography in [10] In particular, Chapter 3 of [10] contains a very nice discussion of region merging segmentation algorithms, starting with Brice and Fennema s [3] and Pavlidis [13] which may be considered as ancestors to both [9] the snakes [8] and our SIDEs. B. Shock Filters and Total Variation. Replacing the discrete vector u(t) with a function u(t, x) of a continuous spatial variable x, and replacing first differences with derivatives in Eq. 4) ....

C. Brice and C. Fennema. Scene analysis using regions. Artificial Intelligence, 1, 1970.

....and analysis. 1. INTRODUCTION The most common way to implicitly represent the regions of a 2D discrete image is the boundary representation. Two main approaches have been used. Roughly speaking, the first one [Fre61] consists in drawing the boundary on the pixels while the second one[BF70] consists in drawing it between the pixels. According to the terminology of inter pixel boundary used by Fiorio[Fio95] we call this second approach the inter pixel oriented approach, and thus the first one the pixel oriented approach. Region boundaries present many interests from the point of ....

....it is generally useless to deal with the contour of the background. On the other hand it is a major drawback when considering an image decomposed into several adjacent objects. An alternative to pixel boundaries is inter pixel boundaries which has been introduced in 1970 by Brice and Fennema [BF70] as a data structure for implementing grouping segmentation. The pixels are considered as squares, two adjacent pixels sharing a vertical or an horizontal edge. A boundary is a sequence of pixel edges encoded by a sequence of points located at pixel corners. From the earlier eighties several ....

C.R. Brice and C.L. Fennema. Scene analysis using regions. Artificial intelligence, 1:205--226, 1970.

....The task of discerning foregriIn biological systems, visio The formulation and implementation of a randomized search approach to segment an image, using genetic algorithms, is presented in this chapter. A classic approach based on state space techniques for segmentation due to Brice and Fennema[2] is reformulated using genetic algorithms. The state space representation of a partially segmented image lends itself to binary strings, in which the dominant substrings are easily explained in terms of chromosomes. Also the operations such as crossover and mutations are easily abstracted. In ....

....Then, one could merge these homogeneous candidate blocks into bigger groups of homogeneous regions. Both the split and merge steps are thus necessary in order to facilitate regions of arbitrary shapes and sizes. Region based techniques using classical state space approach was first applied in[2] and extended in[4] This approach regards the initial image as a discrete state and each pixel as a separate region. Transitions in states take place when a boundary is inserted or removed between any two regions. The problem is thus transformed into searching for allowable changes in the ....

[Article contains additional citation context not shown here]

C. Brice and C. Fennema. Scene Analysis Using Regions. Artificial Intelligence, AI-1(No. 3):pp. 205--226, 1970.

....the probability of a given solution assuming gaussian departures from the model. Leclerc [Lec89] showed that the minimal solution could also be interpretated as the minimal length encoding describing the scene in terms of a given descriptive language. There are stochastic [GG84] region growing [BF70, HP74], and continuation [Lec89, BZ87, GY91] methods for finding solutions to the scene partitioning problem when it is described in terms of a cost functional. Stochastic methods use simulated annealing programs in which a gradient descent method is perturbed by a stochastic process which decreases in ....

C. Brice and C. Fennema. Scene analysis using regions. Artificial Intelligence, 1(3):205--226, 1970.

....for certain applications. In this paper, we discuss the interest of Binary Partition Trees to create new connected operators that do not suffer from this restriction. # Segmentation: A large number of segmentation techniques such as region growing or watershed rely on iterative merging strategies [2], 10] 9] These algorithms sequentially merge either pixels or regions. In practice, the class of rules used to control the merging process is restricted. Indeed, rules involving the global optimization of a criterion that has no specific property (such as increasingness) are not ....

C.R. Brice and C.L. Fenema. Scene analysis using regions. Artificial intelligence, 1:205--226, 1970.

....unnecessary because it was assumed that programs using higher level reasoning would be able to understand, identify, and merge these simple regions as appropriate. The segmentation in computer vision began with the work of Brice Fenema in 1970, who first proposed scene analysis using regions [6]. They modeled images as regions with Gaussian distributions of color and intensity. Based on this approach, Yakimovsky Feldman developed a system for analyzing complex natural scenes [61] Their approach started with unsupervised clustering of pixels followed by a step which merged similar ....

C. R. Brice and C. L. Fenema, "Scene analysis using regions," Artificial Intelligence 1, 1970, pp. 205-226.

....of such objects from a digital image. Some theories propose a segmentation of the image into connected regions by a variational principle [52, 53] Other theories assume that the discontinuity set of the image provides curves which, in some way or another, can be closed by an algorithm (see [8, 50] and the discussion in [7] Canny s filter [9] for instance, computes a set of discontinuity points in the image which must be thereafter connected by some variational principle. The obtained curves are supposed to be the boundaries of the shapes of the image. Many pattern recognition theories ....

C. Brice and C. Fennema. Scene analysis using regions. Artificial Intelligence, 1 (1970), 205--226.

....defined as sequence of pixels is that they are not topologicaly consistent. Moreover two adjacent regions do not share boundary elements. An alternative to overcome these problems is to consider inter pixel boundaries (see Figure 8) This approach have been first described by Brice and Fennema [5] when introducing grouping segmentation algorithm. Later several discrete topologies have been developed which provide formal tools to study such a representation [19] 20] 15] The boundary of a region is made of closed paths of half integer points[12] which are the points of the plane P 1 2 ....

R. Brice and C.L. Fennema. Scene analysis using regions. Artificial intelligence, 1:205--226, 1970.

....finding the region containing a given point. 3. The splitting is the decomposition of a region into sub regions and the merging is the fusion of two or more adjacent regions. Many methods of image segmentation are based on splitting of regions into sub regions [Lee86, BP87] on merging of regions [BF70, BHA81, BGR 89] or on combinations of the both ones [HP76, PR82, CLC91] Remark that the deletion of a region r can be considered as the merging of the region r with an adjacent one. 4. The union of regions can be used in image analysis or in image editing to build complex objects by ....

....two adjacent objects do not share the same boundary element. Moreover such discrete curves do not necessarily satisfy Jordan s theorem. An alternative to pixel boundaries is inter pixel boundaries. An empirical presentation of interpixel boundary have been first given in 1970 by Brice and Fennema [BF70] when introducing merging segmentation. A more theoretical approach have been developed since 1980 introducing discrete topology. The solution generally adopted is to extend the concept of pixel [RK82, KR89, Kov89, Bie90, AAF95, KKM90b, KKM90a, KKM91] to allow a consistent definition of open and ....

C.R. Brice and C.L. Fennema. Scene analysis using regions. Artificial intelligence, 1:205--226, 1970.

....plane when defining discrete curves by any 4 connected (or 8 connected) paths. A path of the discrete plane is a sequence of neighboring points according to 4 or 8 connectivity. A correct topological framework is given by the approach consisting in drawing boundaries between pixels of the scene [11, 15, 1]. We adopt the terminology of Fiorio [18] and call these boundaries inter pixel boundaries. We encode inter pixel boundaries in an alternative discrete plane, that we call the half integer plane. The elements of the half integer plane are the points of the Euclidean plane with coordinates of the ....

C.R. Brice and C.L. Fennema. Scene analysis using regions. Artificial intelligence, 1:205--226, 1970.

....the regions of the image satisfy a given homogeneity criterion. Among region segmentation methods, three main different approaches may be distinguished: i) The top down approach [21,19,8] starts with large regions and divide the image into smaller and smaller regions. ii) The bottom up approach [5,14] starts with small regions and merges them until the homogeneity criterion is satisfied. iii) The mixed approach [6,22,9] introduced by Pavlidis and Horowitz [15,16] modifies the partition by splitting or merging regions until they satisfy some homogeneity criterion. Horowitz and Pavlidis have ....

R. Brice and C.L. Fennema. Scene analysis using regions. Artificial intelligence, 1:205--226, 1970.

....progressively merging regions to create a partition of the image. For instance in the Split Merge [4] algorithm, the set of initial regions is defined by the Split process and the merging is performed between the initial regions depending on a homogeneity criterion. The Region growing algorithm [1] is another example: it relies on the merging of the set of initial regions with individual neighboring pixels that belong to an uncertainty area. Finally, the classical morphological tool for segmentation is the watershed [8] 19] It also relies on a merging strategy: the initial regions are ....

C.R. Brice and C.L. Fenema. Scene analysis using regions. Artificial intelligence, 1:205--226, 1970.

....up will be repeated until for every part of the input space the desired approximation precision is achieved. This process can be compared with the first part of the well known split and merge process as used in image processing. This method used in image processing was first mentioned in [9]. On the other hand we want to have a very flexible system being able to adjust itself to changes in the system. This means that from the moment that the representation as constructed in the tree does no longer match the measured system (i.e. the learning samples are no longer represented by the ....

C. Brice and C. Fennema. Scene analysis using regions. Artificial Intelligence, 1:205--226, 1970.

....criterion computed on one region or on one region and one of its adjacent regions. Among region based algorithms, we distinguish three main approaches: The top down approach [18, 16, 7] starts with an under segmentation of the image and iterates split operations; The bottom up approach [3, 10] starts with an over segmentation of the image and iterates merge operations; The mixed approach [11, 12, 4, 8, 19] modifies the partition by combining split and merge operations. The data structures commonly used by the top down approach are hierarchical structures like quadtrees [9, 20] or ....

....of the adjacency of regions involves complex and costly processing. Another drawback implied by a regular subdivision is the square aspect of the final segmented image. The bottom up approach is frequently implemented by an array of labels [17] combined with a region adjacency graph(RAG) [8, 3]. When the mixed approach is implemented both with regular pyramids and RAG, region merging must preserve the hierarchical structure. In this case, the merging operation is called restricted merge. Due to the incompatibility of data structures used by the split and the merge algorithms, the mixed ....

R. Brice and C.L. Fennema. Scene analysis using regions. Artificial intelligence, 1:205--226, 1970.

....for Finding Regions Regions are areas where the image properties of interest, such as pixel intensity or colour, are uniform. Since uniform regions usually correspond with parts of the surface of the same physical object, region segmentation is often performed as a basis for object recognition. Brice and Fennema (1970) partitioned the image into regions according to pixel intensity thresholds. Regions were then joined together if they shared a common boundary, and there was only a small difference in intensity across this boundary, and certain other heuristic criteria were satisfied. Straight lines were then ....

Brice, C.R. and Fennema, C.L., "Scene analysis using regions," Artificial Intelligence 1, 1970, 205--226.

....and multiassociative techniques yields the best performance in practice. 5. 1 A Segmentation System Many segmentation systems are based either on image splitting guided by the analysis of feature histograms [25, 23] or region merging based on local region properties constrained by global criteria [4, 7, 29]. The system we have used to generate problem instances the Nagin Kohler Griffith Beveridge system combines these two approaches [2] In the first phase, spatially localized feature histograms are used to over segment the image. That is, sensitivity parameters are set so that all region ....

C.R. Brice and C.L. Fennema, "Scene Analysis Using Regions," Artificial Intelligence, Volume 1, pp. 205-226, 1970.

....for this subspace. This process of splitting up will be repeated until for every part of the input space the desired approximation precision is achieved. This process can be compared with the first part of the well known split and merge process as used in image processing, as first mentioned in [1]. A second requirement is that we want to have a very flexible system, able to adjust itself to changes in the system. This means that from the moment that the representation as constructed in the tree does no longer match the measured system (i.e. the learning samples are no longer represented ....

C. Brice and C. Fennema. Scene analysis using regions. Artificial Intelligence, 1:205--226, 1970.

....for this subspace. This process of splitting up will be repeated until for every part of the input space the desired approximation precision is achieved. This process can be compared with the first part of the well known split and merge process as used in image processing, as first mentioned in [1]. A second requirement is that we want to have a very flexible system, able to adjust itself to changes in the system. This means that from the moment that the representation as constructed in the tree does no longer match the measured system (i.e. the learning samples are no longer represented ....

C. Brice and C. Fennema. Scene analysis using regions. Artificial Intelligence, 1:205--226, 1970.

....split and merge operations on regions are called region based methods. Three main region based methods can be distinguished: 1. The top down approach [28, 26, 9] begins with an under partition of the image and increases the number of regions using the split algorithm. 2. The bottom up approach [5, 15] begins with an over partition of the image and decreases the number of regions using the merge algorithm. 3. The mixed approach [6, 10, 29] introduced by Pavlidis and Horowitz [19, 20] modifies the partition combining the split and merge algorithms. The region based methods require different ....

....of the adjacency regions on a tree structure involves complex and costly processing. Moreover, the merge of two adjacency regions may break the tree structure. The usual data structure to implement the merge algorithm is an array of labels [27] combined with a region adjacency graph (RAG) [10, 5]. An array of labels associates to each pixel a label such that all the pixels of a given region have a same label. The vertices of the adjacency graph represents the regions of the image and there exists one edge between two regions if they are adjacent. The merge of two regions consists in ....

R. Brice and C.L. Fennema. Scene analysis using regions. Artificial intelligence, 1:205--226, 1970.

....of the image plane, and not a hierarchical description, the hierarchical approach is very useful. As it is not possible to search the space of all possible image partitions, an efficient search can be made by repeatedly splitting and merging regions. The region merging scheme of Brice and Fennema [15] can be seen as a hierarchical method avant la lettre. In this method, each connected component of the image grid which has a uniform grey value, is used as an initial region. Then regions are merged, until maximal regions are formed which still satisfy a homogeneity criterion. The application of ....

....to evaluate all groupings. Groups of pixels which satisfy the model are typically found by splitting and or merging groups repeatedly until the result fits the model within a given error. Among bottom up grouping schemes, region merging and region growing can be discerned. Region merging methods [15] first consider each pixel as an individual region. Two regions are replaced by their union if the latter satisfies 60 Model Based Bottom Up Grouping the model. Merging continues until no union of adjacent regions satisfies the model. Region growing methods [112] first select a special set of ....

C. R. Brice and C. L. Fennema. Scene analysis using regions. Artifical Intelligence, 1:205--226, 1970.

No context found.

C. Brice and C. Fennema, "Scene analysis using regions," Art. Intell., vol. 1, pp. 205--226, 1970.

No context found.

C. Brice and C. Fennema. Scene analysis using regions. Artificial Intelligence, 1(3):205--226, 1970.

No context found.

C.R. Brice and C.L. Fennema, Scene analysis using regions, Artificial Intelligence, vol. 1, pp. 205-226, 1970.

No context found.

C.R. Brice and C.L. Fennema, Scene analysis using regions, Artificial Intelligence, vol. 1, pp. 205226, 1970.

No context found.

C.R. Brice and C.L. Fennema, Scene analysis using regions, Artificial Intelligence, vol. 1, pp. 205-226, 1970.

No context found.

C. R. Brice and C. L. Fennema. Scene analysis using regions. Artifical Intelligence, 1:205--226, 1970.

No context found.

C. R. Brice and C. L. Fennema. Scene analysis using regions. Arti cial Intelligence, 1(3):205-226, 1970. 17

No context found.

C. Brice and C. Fennema, "Scene analysis using regions," Artificial Intelligence 1(3), pp. 205--226, 1970.

No context found.

C. Brice and C. Fennema. Scene analysis using regions. Artificial Intelligence, 1(3):205--226, 1970.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC