Publication: Evaluation and improvement of a face detection method on the basic of skin color
Loading...
Identifiers
Publication date
2011-09
Defense date
2012-06-22
Authors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Este proyecto aborda el análisis y evaluación de un método de detección de rostros basado en el color de la piel, es decir, teniendo en cuenta principalmente las características de color de la imagen para llevar a cabo la detección de rostros. Este proceso se logra a través de tres pasos fundamentales: la transformación del espacio de color, la generación de imágenes en escala de grises y la estimación del valor umbral óptimo que nos lleva al último paso que es la segmentación. La detección de rostros es un paso crítico en cualquier aplicación de análisis facial, tal como el reconocimiento de rostros, la codificación de vídeo para videoconferencia, inteligencia artificial, etc. En general, este tipo de detección suele implicar problemas muy complejos, ya que los rostros tienen diferentes colores, expresiones, poses y tamaños relativos o se encuentran en diferentes condiciones de iluminación. _________________________________________________________________________________________________________________________
This project deals with the analysis and evaluation of a face detection method based on skin color, ie, taking into account mainly the color characteristics of the image to perform face detection. This process is accomplished through three fundamental steps: transformation of color space, generating grayscale images and establishing optimal thresholding value to reach the last step that is segmentation. To evaluate the proposed algorithm we have decided to use two databases with different characteristics. The first works with images with a controlled background, specifically white and smooth, the only change are the characters of the images and the illumination. The second database works with a not controlled background, that is, each picture is taken in a different environment, with different luminosities, characters and objects, these changes hinder the detection process as discussed below. After analyzing the performance of the face detector we must determine whether the results we have are good for each of the situations that arise
This project deals with the analysis and evaluation of a face detection method based on skin color, ie, taking into account mainly the color characteristics of the image to perform face detection. This process is accomplished through three fundamental steps: transformation of color space, generating grayscale images and establishing optimal thresholding value to reach the last step that is segmentation. To evaluate the proposed algorithm we have decided to use two databases with different characteristics. The first works with images with a controlled background, specifically white and smooth, the only change are the characters of the images and the illumination. The second database works with a not controlled background, that is, each picture is taken in a different environment, with different luminosities, characters and objects, these changes hinder the detection process as discussed below. After analyzing the performance of the face detector we must determine whether the results we have are good for each of the situations that arise
Description
Keywords
Reconocimiento de formas, Proceso de imágenes, Detección de rostros