SAR Image Segmentation With Rényi's Entropy

Shannon Entropy for the $\mathcal {G}^0_I$ Model: A New Segmentation Approach

“…They have obtained good segmentation results, but have indicated the use of numerical integration for entropy as a hard disadvantage of the proposed method. The next result furnishes a solve for the former question we put as well as for the open question in [11].…”

“…g:l,l , by Eqs. (10) and (11). 4: The conditional probability of observationḢ S (i) belong to gth class is calculated byτ g (θ|Ḣ S (i)) in Eq.…”

“…I Model Now, we look for answering about how to obtain a closedform expression for G 0 I distribution entropy and to apply the asymptotic distribution for its stochastic version in processing of SAR intensities. It is important to note Nobre et al [11] have worked with the G 0 A (counterpart of G 0 I for modeling amplitude instead of intensity) Rényi entropy, but without using closedform expressions for the entropy nor working with distributions for entropy. They have obtained good segmentation results, but have indicated the use of numerical integration for entropy as a hard disadvantage of the proposed method.…”

“…) and T g is the pixels numbers at the gth region for an under-study image. Once there are p(p + 1)/2 parameters to be estimated in (11), it is common to impose more parsimony structures in Gaussian mixture models.…”

The Geodesic Distance between $\mathcal {G}_I^0$ Models and its Application to Region Discrimination

“…The problem of comparing two or more samples using distances or divergences, prevails in image processing and analysis. Several authors employ information-theoretic measures of contrast among samples in classification [26], [27], edge detection [28]- [33], saliency detection [34] and despeckling [35]. In the article [36], the authors use the concept of Mutual Information from the Information Theory, to the glacier velocity monitoring by analyzing temporal PolSAR images.…”