Finite state vector quantisation with neural network classification of states

Manikopoulos, C.N.

doi:10.1049/ip-f-2.1993.0022

Cited by 10 publications

(8 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, the memory requirements for DFSVQ-2 are increased by a factor of four in comparison with DFSVQ-1 (about 240 times more than that required by the proposed FSRVQ). Therefore, a small degradation in performance of the proposed FSRVQ scheme can be justified by a savings of [11] AND [12], RESPECTIVELY, FOR THE TEST IMAGE LENA two orders of magnitude in the memory requirements when compared with DFSVQ-2. Table IV shows the performance of the proposed FSRVQ along with the FSVQ schemes developed in [11] and [12], respectively, for the test image Lena.…”

Section: Resultsmentioning

confidence: 99%

“…At the th stage of the th state-RVQ, the error that has to be minimized to obtain the best th reproduction codevector for the respective stage-wise residual input vector is given by (12) The update equation for each weight vector is given by (13) where is the gradient with respect to , or equivalently (14) After differentiating with respect to , we get (15) for . and are defined in a similar manner as and .…”

Section: B Tree-structured Competitive Neural Networkmentioning

confidence: 99%

“…This suggests that the conventional methods, indeed, incorporate a large number of redundant states. Manikopoulos [12] suggested another FSVQ scheme which used the same next-state function as that of Kim's scheme. However, in his scheme a neural-network classifier is used which classifies all the possible states into a small number of representation states.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Finite-state residual vector quantization using a tree-structured competitive neural network

Rizvi

Nasrabadi²

1997

IEEE Trans. Circuits Syst. Video Technol.

View full text Add to dashboard Cite

Finite-state vector quantization (FSVQ) is known to give better performance than the memoryless vector quantization (VQ). This paper presents a new FSVQ scheme, called finite-state residual vector quantization (FSRVQ), in which each state uses a residual vector quantizer (RVQ) to encode the input vector. This scheme differs from the conventional FSVQ in that the state-RVQ codebooks encode the residual vectors instead of the original vectors. A neural network predictor estimates the current block based on the four previously encoded blocks. The predicted vector is then used to identify the current state as well as to generate a residual vector (the difference between the current vector and the predicted vector). This residual vector is encoded using the current state-RVQ codebooks. A major task in designing our proposed FSRVQ is the joint optimization of the next-state codebook and the state-RVQ codebooks. This is achieved by introducing a novel tree-structured competitive neural network in which the first layer implements the next-state function, and each branch of the tree implements the corresponding state-RVQ. A joint training algorithm is also developed that mutually optimizes the next-state and the state-RVQ codebooks for the proposed FSRVQ. Joint optimization of the next-state function and the state-RVQ codebooks eliminates a large number of redundant states in the conventional FSVQ design; consequently, the memory requirements are substantially reduced in the proposed FSRVQ scheme. The proposed FSRVQ can be designed for high bit rates due to its very low memory requirements and the low search complexity of the state-RVQ's. Simulation results show that the proposed FSRVQ scheme outperforms conventional FSVQ schemes both in terms of memory requirements and the visual quality of the reconstructed image. The proposed FSRVQ scheme also outperforms JPEG (the current standard for still image compression) at low bit rates.Index Terms-Competitive neural network, image coding, joint optimization, multilayer perceptron, vector quantization.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: B Tree-structured Competitive Neural Networkmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Finite-state residual vector quantization using a tree-structured competitive neural network

Rizvi

Nasrabadi²

1997

IEEE Trans. Circuits Syst. Video Technol.

View full text Add to dashboard Cite

show abstract

“…Properties such as topology preserving, flexibility, potential as a robust substitute for clustering and visualization analysis [25], learning ability from complex, multi-dimensional data and transformation to visual clusters [26] make a SOM to a very efficient tool for many applications. Some examples are: Yardangs idenfication [27], image data compression [28], image or character recognition [29,30], robot control [31,32], lithological discrimination using Landsat TM [33], urban land use classification [34], ecological modeling [22], landscape elements analysis [35], visualizing high-dimensional data [36], mapping surface geology [37], financial markets [38,39], data mining and knowledge discovery [40][41][42], hyperspectral image classification [43][44][45] and classification of remote sensing data [46,47].…”

Section: Introductionmentioning

confidence: 99%

Effect of SRTM resolution on morphometric feature identification using neural network—self organizing map

2009

View full text Add to dashboard Cite

In this study, we present a semi-automatic procedure using Neural NetworksSelf Organizing Map-and Shuttle Radar Topography Mission DEMs to characterize morphometric features of the landscape in the Man and Biosphere Reserve "Eastern Carpathians". We investigate specially the effect of two resolutions, SIR-C with 3 arc seconds and X-SAR with 1 arc second for morphometric feature identification. Specifically we investigate how the SRTM/C band data with 30 m interpolated grid, corresponding to SRTM/X band 30 m, affect the morphometric characterization and topography derivatives. To reduce misregistration between the DEMs, spatial co-registration was performed and a RMSE of 0.48 pixel was achieved. Morphometric parameters such as slope, maximum curvature, minimum curvature and cross-sectional curvature are derived using a bivariate quadratic approximation on 90 m, 30 m and interpolated 30 m DEMs. Self Organizing Map (SOM) is used for the classification of morphometric parameters into ten exclusive and exhaustive classes. These classes were analyzed as morphometric features such as ridge, channel, crest line and planar for all data sets based on feature space (scatter plot), morphometric signatures and 3D inspection of the area. The map quality is analyzed by oblique views with contour lines overlaid. Using the X band DEM with 30 m grid as benchmark, a change detection technique was used to quantify differences in morphometric features and to assess the scale effect going from a 90 m (C-band) DEM to an interpolated Geoinformatica (2010) 14:405-424 30 m DEM. The same procedure is used to study the effect of different resolutions on morphometric features. Morphometric parameters were computed by a moving window size 5×5 (corresponding to 450 m on the ground) over SRTM-90 m. To cover the same ground area, a moving window size of 15×15 is used for the 30 m DEM. The change analysis showed the amount of resolution dependency of morphometric features. Overall, the results showed that the introduced method is very useful for identification of morphometric features based on SRTM resolution. Decreasing the grid size from 90 m to 30 m reveals considerably more detailed information emphasizing local conditions. Comparison between results from DEM-30 m as reference data set and interpolated 30 m, showed a rate of change of 31.5% which is negligible. About 17% of this rate correspond to classes with mean slope>10°. Of the morphometric parameters, the cross sectional curvature is most sensitive to DEM resolution. Increasing spatial resolution reduces the main constrains for morphometric analysis with SRTM 90 m data, such as unrealistic features and isolated single elements in the output map. So in case of lack of high resolution data, the SRTM 90 m data could be interpolated and used for further geomorphic analysis.

show abstract

“…Such types of topological structures would also denote the characteristic of the input samples. SOM has been employed in a wide range of applications in various different domains, including speech recognition, image data compression, robot control, pattern recognition, and medical diagnosis [13], [14], [17], [18], [20], [22], [24], [26], [27].…”

Section: Introductionmentioning

confidence: 99%

A New Clustering Validity Index for Cluster Analysis Based on a Two-Level SOM

Shieh

Liao

2009

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

SUMMARYSelf-Organizing Map (SOM) is a powerful tool for the exploratory of clustering methods. Clustering is the most important task in unsupervised learning and clustering validity is a major issue in cluster analysis. In this paper, a new clustering validity index is proposed to generate the clustering result of a two-level SOM. This is performed by using the separation rate of inter-cluster, the relative density of inter-cluster, and the cohesion rate of intra-cluster. The clustering validity index is proposed to find the optimal numbers of clusters and determine which two neighboring clusters can be merged in a hierarchical clustering of a two-level SOM. Experiments show that, the proposed algorithm is able to cluster data more accurately than the classical clustering algorithms which is based on a two-level SOM and is better able to find an optimal number of clusters by maximizing the clustering validity index.

show abstract

Finite state vector quantisation with neural network classification of states

Cited by 10 publications

References 2 publications

Finite-state residual vector quantization using a tree-structured competitive neural network

Finite-state residual vector quantization using a tree-structured competitive neural network

Effect of SRTM resolution on morphometric feature identification using neural network—self organizing map

A New Clustering Validity Index for Cluster Analysis Based on a Two-Level SOM

Contact Info

Product

Resources

About