Image classification based on color and texture analysis
Begoña Acha, Carmen Serrano · 2002
We propose a method to classify color images into different groups based on texture and color information. We take advantage of color information by clustering with a vector quantizer algorithm the color centroid of each image into the desired groups in the plane (V/sub 1/, V/sub 2/), the two chrominance components of the CIE Lab representation. To decrease the number of images misclassified, we then apply a texture analysis to the images; specifically, we calculate the statistical parameters, kurtosis and skewness. We apply the whole procedure to classify burn wound images. It was tested with 80 images, classifying without failure 88.75% of them.