Discrete optimal filtering (Q4522767)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 1548580
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | Discrete optimal filtering |
scientific article; zbMATH DE number 1548580 |
Statements
Discrete optimal filtering (English)
0 references
14 August 2001
0 references
binary signal
0 references
noise
0 references
addition modulo 2
0 references
contrast
0 references
noisy pictures decoding
0 references
0.9158838
0 references
0 references
This paper is about a process \(\{\underline W_i,\;i\in [1:n]\}\) built as follows: \[ \underline W_i= {U_i\brack V_i},\quad 1\leq i\leq n, \] where 1. \(\{U_i,\;i\in [1:n]\}\) is a given (deterministic) unobservable sequence of zeros and ones which contains at least one sequence of zeros and one sequence of ones, both of length \(p\). 2. \(\{V_i=U_i \oplus B_i,\;1\leq i\leq n\}\), where \(\oplus\) is addition modulo 2. There is an integer \(q\) such that the noise process \(\{B_i,\;i\in [1:n]\}\) has the following property: it is a (random) sequence of zeros and ones such that for every fixed, but arbitrary \(m\in [1:p \wedge q]\), the law of \(\{B_i,\;i\in [1:n]\}\) restricted to \([i:i+m-1]\), denoted \(\underline B[i:i+ m-1]\), is independent of \(i\). One is looking for functions \[ w(\varepsilon_1, \dots,\varepsilon_m) =w(\underline \varepsilon_m) \in\{0,1\}, \quad\varepsilon_i \in\{0,1\},\;1\leq i\leq m, \] such that \(w\) maximizes a contrast defined as \[ \bigl|\mu^{(m)}(w) \bigr|= \left|\sum_{\{\underline \varepsilon_m\}} w(\underline \varepsilon_m) \bigl\{p^{(m)} (\underline \varepsilon_m \oplus \underline 1_m)-p^{(m)} (\underline \varepsilon_m) \bigr\} \right|, \] where \(\underline \varepsilon_m \oplus \underline 1_m\) has components \(\varepsilon_i \oplus 1\), \(1\leq i\leq m\), and \[ p^{(m)} (\underline \varepsilon_m)= P\bigl(\underline B[i:i+m-1]=\underline \varepsilon_m \bigr). \] If \(w^*\) denotes the optimal \(w\), the sequence \(\{U_i,\;i\in[1:n]\}\) is then reconstructed as the sequence \(\{w^* (\underline V[i:i+m-1])\), \(1\leq i\leq n-m+1\}\). The paper exhibits a method to obtain \(w^*\), contains a formula that yields the maximum value of the contrast in terms of the noise probabilities, explains how the estimation procedure just described is equivalent to a testing problem, and shows how it can be used to decode a noisy, black-and-white encoded picture.
0 references