10.03.2015 Views

sparse image representation via combined transforms - Convex ...

sparse image representation via combined transforms - Convex ...

sparse image representation via combined transforms - Convex ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

2.2. SPARSITY AND COMPRESSION 17<br />

with the largest n amplitudes, then the compression number c(n) is the square root of the<br />

RSS distortion of the signal reconstructed by the coefficients with the n largest amplitudes.<br />

The following result can be found in [47].<br />

Lemma 2.2 For any sequence θ, ifm =1/p − 1/2, the following inequality is true:<br />

c(N) ≤ α p N −m |θ| wl p, N ≥ 1,<br />

where α p is a constant determined only by the value of p.<br />

From the above lemma, a small weak l p norm implies a small compression number.<br />

Rate of Recovery<br />

The rate of recovery comes from statistics, particularly in density estimation. For a sequence<br />

θ, the rate of recovery is defined as<br />

r(ɛ) =<br />

∞∑<br />

min{θi 2 ,ɛ 2 }.<br />

i=1<br />

Lemma 2.3 For any sequence θ, ifr =1− p/2, the following inequality is true:<br />

r(ɛ) ≤ α ′ p|θ| p wl p (ɛ 2 ) r , ɛ > 0,<br />

where α ′ p is a constant.<br />

This implies that a small weak l p norm leads to a small rate of recovery. In some cases<br />

(for example, in density estimation) we choose rate of recovery as a measure of sparsity.<br />

The weak l p norm is therefore a good measure of sparsity too.<br />

Lemma 1 in [46] shows that all these measures are equivalent in an asymptotic sense.<br />

Critical Index<br />

In order to define the critical index of a functional space, we need to introduce some new<br />

notation. A detailed discussion of this can be found in [47]. Suppose Θ is the functional<br />

space that we are considering. (In the transform coding scenario, the functional space Θ<br />

includes all the coefficient vectors.) An infinite-length sequence θ = {θ i : i ∈ N} is in a weak

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!