PILAR ROSADO
ART & RESEARCH 

VISUAL
VOCABULARIES

The struggle between writing and image is a secular one. Today, there are numerous indications that a return to the image is necessary: it is important to find analogies with language that can be applied to visual information. If it has been possible to break down language into elements and structures, would it also be possible to do so with images? Digital imaging offers us the opportunity to describe forms in mathematical terms, thus providing us with the means to decipher the problem of the meaning contained in an image.
  
Read more

An image description algorithm used in computer vision has been programmed, whose approach consists of placing a regular grid of interest points on the image and selecting a region of pixels around each of its nodes, for which a descriptor is calculated that takes into account the gradients of gray found. The descriptors of the entire collection of images can be grouped based on their similarity, and each resulting group will determine what we call a "visual word." The total of "visual words" in the image collection generates a visual vocabulary. The method is called Bag-of-Words because it represents an image as a disordered collection of local visual features. Considering the frequency with which each "visual word" occurs in each image, we apply pLSA (Probabilistic Latent Semantic Analysis), a statistical model that will automatically classify images according to their formal category.
2018