Mid-Level Image Descriptors

Mid-Level Image Descriptors

Jean Martinet (University of Lille, France) and Ismail Elsayad (University of Lille, France)
DOI: 10.4018/978-1-61350-126-9.ch003
OnDemand PDF Download:
List Price: $37.50


We present in this chapter a classification of image descriptors, from the low level to the high level, introducing the notion of intermediate level. This level denotes a representation level lying between low-level features – such as color histograms, texture or shape descriptors, and high-level features – semantic concepts. In a chain of process point of view, mid-level descriptors represent an intermediate step or stage between low and high level, dedicated to specific tasks such as annotation, object detection/recognition, or similarity matching. After introducing a definition for the three different levels, we review a number of approaches making use of such intermediate levels. We namely focus on different approaches making an analogy with text processing, by adapting and applying standard text processing techniques to image indexing.
Chapter Preview


The notion of intermediate level (or mid-level) descriptor is not new. For instance, Luo & Savakis (2001) proposed the use of a Bayesian network for integrating knowledge from low-level to mid-level features for indoor/outdoor classification of images. The network integrates low-level features (color and texture) and so-called mid-level features (external knowledge about sky and grass) using a single classification engine. Mylonas et al. (2007) have also used some mid-level descriptors. Their work aims at improving both image segmentation and labeling of materials and simple objects at the same time, with obvious benefits for problems in the area of image understanding. The novelty of the proposed idea lies on blending well-established segmentation techniques with mid-level features. ARG (Attributed Relational Graphs) were used as mid-level descriptors in their approach, so that images can be described as structured sets of individual objects, thus allowing a straightforward mapping to a graph structure. We provide below process-oriented definitions of low-, high-, and mid-level descriptors. Figure 1 shows a schematic description of the situation each level of description.

Figure 1.

Schematic description of the situation of mid-level descriptors

Complete Chapter List

Search this Book: