| Summary: | 博士 === 國立臺灣科技大學 === 電機工程系 === 102 === With advances in technology, image-processing techniques are changing daily. All high tech products are moving toward controls using human touch. Therefore, In the future, human skin color detection will have an important role in applications such as skin color recognition and image database management, human computer interfaces and video surveillance. This study has three parts: (1) a comparison of the effect on skin color of the and threshold values for three different races’ skin colors; (2) the segmentation of skin and skin-like objects and (3) the way in which image merging techniques effectively filter out skin-like interference in a skin-like environment and repair broken images.
This study proposes the use of original RGB images to obtain binary images using linear and nonlinear fixed threshold methods, without a need for color space transformation.
This study firstly uses a probabilistic neural network to categorize hand gestures into eight different types and conducts experiments using Hu and contour methods. It is discovered that the contour sequence moment is faster by about 0.001 to 0.005 seconds, for all types of hand gesture calculation.
These descriptions illustrate that the three calculations proposed in this study are proven to be better methods.
|
|---|
