Abstract | ||
---|---|---|
When using a regularized approach for image restora-tionthere is always a compromise between image sharpnessand noise suppression. Therefore, the main problem is toremove as much noise as possible while preserving sharpnessin the restoration. To this cause we introduce a spa-tiallyregularized neural approach that makes use of localimage statistics to apply varying regularization to differentareas of the image. This is achieved with an efficient parallelimplementation of the Hopfield neural network. Theproposed approach exhibits an improvement in restorationquality and execution time over the existing approaches.This is illustrated on simulations on benchmark images. |
Year | DOI | Venue |
---|---|---|
2002 | 10.1109/SBRN.2002.1181467 | SBRN |
Keywords | Field | DocType |
execution time,hopfield neural network,benchmark image,image restora-tionthere,existing approach,theproposed approach,spatially adaptive image,efficient parallelimplementation,spa-tiallyregularized neural approach,regularized approach,neural network filtering,image sharpnessand noise suppression,scanning electron microscopy,statistics,neural networks,gaussian noise,adaptive filters,magnetic force microscopy,image restoration,filtering,degradation,neural network | Computer vision,Noise suppression,Pattern recognition,Computer science,Filter (signal processing),Regularization (mathematics),Artificial intelligence,Execution time,Image restoration,Filtering theory,Artificial neural network,Machine learning | Conference |
ISBN | Citations | PageRank |
0-7695-1709-9 | 3 | 0.43 |
References | Authors | |
7 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Alex S. Palmer | 1 | 3 | 0.76 |
Moe Razaz | 2 | 23 | 5.63 |
Danilo Mandic | 3 | 1641 | 173.32 |